Resolving similar entities from a database.

0001Embodiments of the invention relate generally to data analysis, and more particularly a dissolution similar entities from a transaction database.

0002A procurement of relevant information from large databases can be relatively straightforward in some situations. Specifically, when the records are in a database well structured and is aimed to obtain information in records that have a specific value or a specific string in a specific field, these records can be isolated using filter functions of database interface software. Using combinations of filter functions, the way to be identified in the records for the isolation, are technically refined. The isolated data sets can then be aggregated to provide a report containing all the records, which together form the desired information.

0003However, support such filtering functions to designate similarities containing database records, attributes that are identical across this database records. In the real world database records may not have attributes that are across the same about these records, although these records relate to each other, or they may have identical attributes in a relatively small number of fields (or portions of fields), such that filter functions not be able to provide an isolation of the desired database records from other database records. For example, such problems occur when a database has database records that originate from several different sources. Isolating of related database records from other database records is a technical problem (which has billions of database records such. As a database) with increasing size of the database, in terms of the number of existing records will be more severe. Since the sizes of databases in the real world is increasing over time, it is expected that this problem is more severe with time.

0004The methods described here can for example be found in the field of financial institutions use, store the transaction data for analysis. A financial institution generates transaction data from credit and debit cards performed purchase transactions with companies that have a merchant account with the financial institution. The merchant account can be used to process individual performed with credit or debit card purchase transactions. Again, is stored as a transaction record in a transaction database, each such purchase. A transaction record, which belongs to a specific merchant account, often involves a merchant ID attribute that associates the transaction record with the merchant account. A Merchant ID can be any data type, including a number, a string, or any combination thereof. The financial institution can then analyze the transaction records of one or more merchant accounts. For example, an analysis include an aggregation of the transaction records of a merchant account or specific merchant accounts. The analysis can then compare the performance of the merchant account with those accounts competing distributor in the same geographical area.

0005Although the financial institution stores the transaction records in a transaction database, may require some analysis that the data is organized in ways that are not part of the transaction records in the database. These databases contain transaction record phrases that would group together an analysis, although there is no single attribute value that sets the transaction records in relationship. For example, if a financial institution is a transaction database configured with a merchant id attribute that associates each transaction record with a merchant account, then you could aggregate with the same merchant id analysis readily transaction records. However, a single company can have multiple merchant accounts at a financial institution. If the financial institution offers different merchant IDs for each merchant account, even if multiple merchant accounts belonging to a single company, it is difficult to aggregate transaction records from multiple merchant accounts this company. For example, a franchise company have different merchant accounts with different merchant IDs for each franchisee location. In such a case, an analysis would not be able to aggregate the transaction records of the franchise company solely on the basis of identical merchant IDs. Instead, an analysis utilize similarities between the merchant id attribute values ​​to aggregate the transaction records of the franchise company.

0006Existing methods based on simple tests, such as comparisons of strings between an attribute in a database of transaction records to detect similarities between groups of transaction records. Transaction records, include the attribute strings which satisfy a measure of similarity, then theAnalysis aggregated. These methods may work, if the attribute contains the same or similar character strings, for groups of transaction records that should be aggregated, and if it contains different strings, for groups of transaction records that should not be aggregated.

0007However, such identifiers are not always (or even not even usually) available. For example, can prevent an analysis system can aggregate the transaction records of the company, different merchant IDs for merchant accounts to a single company. In addition, transaction records may contain similar identifiers, on the basis of an analysis system can perform aggregation, even if the transaction records should not be aggregated. For example, two different companies have merchant accounts with similar merchant IDs that could be attributed to a single company, an analysis system incorrectly. The analysis system may then erroneously aggregate the transaction records of the two companies.

0008Like the previously illustrated Described, there remains a need for more effective methods for evaluating financial transaction records.

0009Embodiments of the invention address the problem of data related database records to identify who may have no usable identical attributes, and thereby exclude unrelated database records, and solve in particular the problem, identify database records that relate to a common entity are, however, may not have identical attributes.

0010The present invention is set out in the independent claims. The dependent claims concern optional features of some embodiments of the invention.

0011An embodiment of the invention sets forth a method for identifying related transaction records from a database, the transaction records for multiple entities stores, which includes: grouping of transaction records with a common attribute value to transaction record sets, accepting a selection of a sample data-set, and determining the likelihood that the transaction record set stores transaction records that are associated with a first entity. The method also includes a dissolution that the transaction record set stores transaction records that are associated with the first entity.

0012include other embodiments of the invention, without limitation, a computer readable storage medium comprising instructions which, when executed by a processing unit cause the processing unit to implement aspects of the approach described herein, as well as a system that includes different elements that configure are to implement aspects of the approach described here.

0013An advantage of the disclosed method is that two data sets, however, belong to a database of transaction records that do not have identical attributes to the same common entity that can be linked to the common entity. Therefore, resolutions that would missed alone with comparing strings, accomplished, and incorrect resolutions that merely based on similar strings are avoided, which improves the precision of a resolution. Another advantage of the disclosed method is that it reduces the number of erroneous aggregations resulting from transaction records, which have similar identifiers, even though they are different entities belonging. By reducing the number of false aggregations is of a delivered in this manner aggregated reporting of data sets less than the memory occupied by a corresponding aggregate message that is generated by means of filter functions.

0014Thus, the previously described features of the invention can be understood in detail, a more particular description of the invention summarized briefly above will be given with reference to embodiments, some of which are illustrated in the accompanying drawings. It is understood, however, that the appended drawings illustrate only typical embodiments of the invention and are therefore not to be understood as limiting the scope, since the invention allows additional equally effective embodiments.

0015<figref>1</figref> is a block diagram illustrating a computer system that is configured to implement one or more aspects of the invention.

0016<figref>2</figref> is a block diagram of the data flow in the application server.

0017<figref>3</figref> shows a method for training of the classifier, according to an embodiment.

0018<figref>4</figref> shows a method for dissolving a merchant ID to a merchant, according to an embodiment.

0019<figref>5</figref> shows an example of a computing environment according to an embodiment.

0020Embodiments of the invention can be used to aggregate certain financial transaction records that resolve to a single entity, but may otherwise could not be grouped together. Assuming that a transaction database of a financial institution identified transaction records of each different merchant account a company using a different merchant id attribute, then it is possible that the different merchant id attributes do not fit together correctly, the transaction records of accounts to link with the company. As another example, have different franchisees a common franchisor to aggregate a flood of merchant accounts, making it difficult, the transaction records that are all franchisees of the franchisor belong solely from the transaction records. In one embodiment, a financial analysis system combines transaction records to merchant id-sets based on identical merchant IDs so that each merchant id-set contains all transaction records with a specific merchant id. How is this example, a single company can be represented by several merchant IDs. To evaluate the complete set of transaction records for a single entity (company), each collection of financial transaction records must (the merchant id phrases) that are of the individual entity associated, are merged.

0021In one embodiment, the analysis system transaction records aggregated from a large collection of merchant id-sets. This aggregation may include that the average transaction size, the standard deviation of the transaction size or the average amount is calculated to an individual has spent. The analysis system uses the aggregations to train a classifier. Once the analysis system has been trained, it creates a confidence whether two merchant id phrases belong to a company, based on aggregations of the pair of merchant id-sets. To associate the merchant id phrases with the company, the analysis system receives a selection of a sample merchant id-set that should be associated with the company and represents the characteristics of the company as possible. The analysis system comparing the Example Merchant ID-set with another merchant id-sets to determine a confidence. The confidence value represents the probability that the sample merchant id-set and the other merchant id-set associated with the company. The analysis system associated each merchant id-set which, has in comparison with the example, confidence values ​​above a threshold value, with the company. This process results in a collection of financial transaction records, to accept of which is that they all belong to a single company, despite the fact that many such records may contain different merchant IDs.

0022In the following description numerous specific details are set forth in order to provide a fundamental understanding of the invention. However, it will be understood by those skilled that the present invention without one or more of these specific details may be executed.

0023<figref>1</figref> is a block diagram showing an exemplary data analysis system <figref>100</figref> represents, according to an embodiment of the invention. As shown, the data analysis system includes<figref>100</figref> an application server <figref>140</figref>Running on a server computer system <figref>130</figref> running a client <figref>120</figref>Running on a client computer system <figref>110</figref> running, and at least one transaction database <figref>160</figref>, Next, the client can<figref>120</figref>, The application server <figref>140</figref> and the transaction database <figref>160</figref> via a network <figref>180</figref> communicate.

0024The client <figref>120</figref> represents one or more software applications that are configured to present data and user input translated to requests for data analysis by the application server <figref>140</figref>, In this embodiment, the client<figref>120</figref> connect to the application server <figref>140</figref>, However, it is also possible that multiple clients<figref>120</figref> on the client computer <figref>110</figref> be executed, or more clients <figref>120</figref> on multiple client computers <figref>110</figref> can use the application server <figref>140</figref> to interact. In one embodiment, the client<figref>120</figref> be a browser that accesses a Web service.

0025Alternatively, the client <figref>120</figref> on the same server computer system <figref>130</figref> as the application server <figref>140</figref> to run. In any case, a user would on the client<figref>120</figref> with the data analysis system <figref>100</figref> to interact.

0026The application server <figref>140</figref> is configured to dealer-resolution tool <figref>150</figref> and an analysis engine <figref>155</figref> includes. The merchant-resolution tool<figref>150</figref> associated matching merchant IDs with a company. The merchant-resolution tool<figref>150</figref> reads data from the transaction database <figref>160</figref>, The merchant-resolution tool<figref>150</figref> can resolution data on the server computer <figref>130</figref> or in the transaction database <figref>160</figref> save.

0027The analysis engine <figref>155</figref> uses the resolution information from the merchant-resolution tool <figref>150</figref>. order from the transaction database <figref>160</figref> analyze data retrieved. The analysis engine<figref>155</figref> introduces aggregating and comparing the transaction records from the transaction database <figref>160</figref> through to provide insights on a specific company. For example, a financial institution to design a data analysis to evaluate the seasonal spending trends for a franchise company. However, each franchisee the franchise company may have a different merchant account with the financial institution. The financial institution stores the transaction records of the merchant accounts with different merchant IDs, which associate a transaction record with a merchant account. To evaluate the complete set of transaction records for the franchise company, it is necessary that the analytical engine<figref>155</figref> any collection of financial transaction records by each franchisee brings together. For this purpose uses the analytical engine<figref>155</figref> the resolution information of the merchant-resolution tool <figref>150</figref>To bring together the financial transaction records of each franchisee to a full set of transaction records for the franchise company to evaluate the seasonal spending trends for the franchise company.

0028In this embodiment, the transaction database stores <figref>160</figref> Records of financial transactions that have a financial institution belongs. For example, the transaction database includes records for a large number of merchant accounts to process credit and debit card transactions. In such a case, each record would contain data attributes for the outstanding amount, the date and time of the transaction, the address of the seller, and a merchant ID to associate the record with a specific merchant account.

0029The transaction database <figref>160</figref> may be a relational database management system (RDBMS) that stores the transaction data as rows in relational tables. Alternatively, the transaction database<figref>160</figref> on the same server computer system <figref>130</figref> as the application server <figref>140</figref> be saved. The records of a financial institution [still missing text in the English original]

0030<figref>2</figref> shows a flow of data from the transaction database <figref>160</figref> by the merchant-resolution tool <figref>150</figref>According to one embodiment of the invention. As illustrated, includes the transaction database<figref>160</figref> Merchant ID phrases <figref>210</figref>, Each merchant id set<figref>210</figref> contains transaction records <figref>215</figref> with the same merchant id, such as credit and debit card transactions are processed for a single merchant account at a financial institution. The merchant-resolution tool<figref>150</figref> includes a Aggregierer <figref>240</figref>, Candidate aggregations <figref>242</figref>, Sample aggregations <figref>244</figref>, A training data set <figref>260</figref> and a Identitätsauflöser <figref>250</figref>, The Identitätsauflöser<figref>250</figref> even includes a classifier <figref>255</figref> and resolution list <figref>270</figref>,

0031In one embodiment, the classifier is <figref>255</figref> a Random Forest classifier. A Random Forest classifier is a machine learning algorithm of the commonly known, that it is extremely accurate in large databases containing discrete, coherent and missing data, as for financial transaction records<figref>215</figref> in the transaction database <figref>160</figref> may be the case. Random Forest classifier include several decision trees. The decision trees to evaluate characteristics of the input data. In the present context of financial transaction records that are associated with merchant accounts via a merchant ID, the evaluated features can include:<ul list-style="bullet"><li>- Wortüberschneidungszählwert and frequency of merchant id attributes</li><li>- Word-based cosine similarity, weighted based on the values ​​related to term frequency inverse document frequency of merchant id attributes</li><li>- Character-based cosine similarity of merchant id attributes</li><li>- Placement of a word Intersection of merchant id attributes</li><li>- Identify the character string ".com"</li><li>- Whether the merchant id attributes include a store code</li><li>- Intersection of prefix or suffix numbers in the merchant id attributes</li><li>- Whether the supplied city code is numerically</li><li>- Matching unique classification code</li><li>- Fractional difference in average transaction size</li><li>- Standard deviation of the average transaction size</li><li>- Fractional difference in the size of the transaction size variances</li></ul>

0032Note that the classifier <figref>255</figref> may evaluate a variety of other characteristics, depending on the requirements of a particular case and of data available from the underlying transaction records. Further recognizes a skilled person that a Random Forest classifier is used as a reference example of a classifier, and that a variety of other machine learning classifiers could be used.

0033To evaluate the diversity of characteristics is in the classifier <figref>255</figref> a growth of decision trees carried out based on the probability that a selected feature to should lead to a certain classification. In the present context the classifier<figref>255</figref> a convergence of several decision trees carried out based on different combinations of features, so that each decision tree classifies that a pair of merchant id-sets <figref>210</figref> fits within the same company, or not. The output value of the classifier<figref>255</figref> is the percentage of decision trees that classify that a pair of merchant id-sets <figref>210</figref> fits within the same company.

0034To prepare to link merchant IDs with a company that is in the classifier <figref>255</figref> a growth of decision trees conducted by exercising on Trainingsdatendatz <figref>260</figref>, The training data set<figref>260</figref> includes pairs of merchant id-sets <figref>210</figref>That fit within the same company, as well as pairs of merchant id-sets <figref>210</figref>That do not fit within the same company. The pairs of merchant id-sets<figref>210</figref>That fit within the same company, are in the training data set <figref>260</figref> classified as positive examples. The pairs of merchant id-sets<figref>210</figref>That do not fit within the same company, are in the training data set <figref>260</figref> classified as negative examples.

0035During the classifier <figref>255</figref> the characteristics of each pair of merchant id-sets <figref>210</figref> processed as a positive or negative example, the accuracy of the classifier is <figref>255</figref> enlarged, in that the probabilities used in the decision trees are refined.

0036The training data set <figref>260</figref> may include difficult borderline cases, for example, pairs of merchant id-sets <figref>210</figref>Which do not match, however, have similar merchant id strings. A pair of merchant id-sets<figref>210</figref> with similar merchant id strings that should not be linked to the same company, is a borderline case because often similar merchant id strings of merchant id-sets <figref>210</figref> come, which should be linked to the same company. Adding such borderline cases for training data set<figref>260</figref> causes the classifier <figref>255</figref> the probabilities in the decision trees of the classifier <figref>255</figref> adapts to better classify pairs of merchant id-sets <figref>210</figref> perform related merchant id attributes.

0037To a large training data set <figref>260</figref> to produce, can the merchant-resolution tool <figref>150</figref> Pairs of randomly selected merchant id-sets <figref>210</figref> produce, deliver the typically negative training examples.

0038The training data set <figref>260</figref> can transaction records <figref>215</figref> include that from the transaction database <figref>160</figref> were read out, can artificial transaction records <figref>215</figref> contain or may contain any combination of these. Although it has become a training data set<figref>260</figref> 4000 pairs of merchant id-sets <figref>210</figref> shown to be effective, however, the actual size of the training data set <figref>260</figref> depending on your preference be determined.

0039Once the classifier <figref>255</figref> has been trained, the merchant-resolution tool <figref>150</figref> be used to associate merchant IDs from a different merchant account with a company, so that the analytical engine <figref>155</figref> Data analysis with full sets of transaction records <figref>215</figref> can perform all merchant accounts to the company.

0040The transaction database <figref>160</figref> is configured such that it includes a mechanism for transaction records <figref>215</figref> with a common merchant id attribute as merchant id phrases <figref>210</figref> supplies. For example, the transaction database<figref>160</figref> Transaction records <figref>215</figref> with same merchant id attributes together in merchant id-sets <figref>210</figref> save, or the transaction database <figref>160</figref> can transaction records <figref>215</figref> sequentially store based on the value of a transaction data attribute. Regardless of the arrangement of the transaction records<figref>215</figref> can the merchant resolution tool <figref>150</figref> Merchant ID phrases <figref>210</figref> from the transaction database <figref>160</figref> recall.

0041After a user a merchant id set <figref>210</figref> as Example Merchant ID Set <figref>210 (0)</figref> has chosen to further merchant id phrases <figref>210</figref> as candidates Merchant ID phrases <figref>210 (1)</figref> to <figref>210 (M - 1)</figref> be considered. The user selects the Example Merchant ID Set<figref>210 (0)</figref> as representative of the characteristics of the company, which is to be dissolved. The Example Merchant ID Set a large number of transaction records<figref>215</figref> include. A large number of transaction records<figref>215</figref> can provide aggregations, for example, the average transaction size, the more accurate than merchant id phrases <figref>210</figref> with a smaller number of transaction records <figref>215</figref> are. Other factors, such as geographical locations, the merchant ID string, or further business heuristics can also select the sample merchant id-set<figref>210 (0)</figref> from the available merchant id-sets <figref>210</figref> to guide.

0042If merchant IDs are associated with a company that calls the merchant resolution tool <figref>150</figref> the transaction records <figref>215</figref> of Example Merchant ID-set <figref>210 (0)</figref> and the transaction records <figref>215</figref> a candidate merchant id-set <figref>210 (1)</figref> from. The Aggregierer<figref>240</figref> aggregating the attributes of the transaction records <figref>215</figref> of Example Merchant ID-set <figref>210 (0)</figref>To sample aggregations <figref>244</figref> to create. For example, the calculatedAggregierer <figref>240</figref> the average transaction size, the standard deviation of the transaction size or the average amount that an individual has spent. The merchant id attribute of Example Merchant ID-set<figref>210 (0)</figref> is also in the sample aggregations <figref>244</figref> contain. The Aggregierer<figref>240</figref> also calculates the candidate aggregations <figref>242</figref> from the candidate merchant id set <figref>210</figref> and also includes the merchant id attribute of the candidate merchant id-set <figref>210 (1)</figref> with the candidate aggregations <figref>242</figref> on. It should be noted that the Aggregierer<figref>240</figref> can calculate additional Aggregierungswerte, according to numerous different designs, which can select the tool developers.

0043After the Aggregierer <figref>240</figref> the Aggregierungswerte has determined directs the merchant-resolution tool <figref>150</figref> the sample aggregation <figref>244</figref> and the candidate aggregation <figref>242</figref> to a Identitätsauflöser <figref>250</figref> continue. The classifier<figref>255</figref> determines the values ​​that are used as features in the decision tree, from the data in the sample aggregations <figref>244</figref> and the candidate aggregations <figref>242</figref> are included. The classifier<figref>255</figref> processes the sample aggregation <figref>244</figref> and the candidate aggregation <figref>242</figref>To generate a confidence between zero and one, which corresponds to how likely the Example Merchant ID Set <figref>210 (0)</figref> for candidates Merchant ID Set <figref>201 (1)</figref> fits, and should therefore be associated with the same company. If the Example Merchant ID Set<figref>210 (0)</figref> and the candidate merchant id set <figref>210 (1)</figref> a value above a certain threshold value obtained, for example, 0.70, then stores the Identitätsauflöser <figref>250</figref> Merchant ID of the candidate merchant id-set <figref>201 (1)</figref> in a resolution list <figref>270</figref>,

0044The merchant-resolution tool <figref>150</figref> compares candidates Merchant ID phrases <figref>210 (2)</figref> to <figref>210 (M - 1)</figref> with a sample merchant id set <figref>210 (0)</figref>, The Identitätsauflöser<figref>250</figref> adds the Merchant ID of each candidate merchant id-set <figref>210 (1)</figref> to <figref>210 (M - 1)</figref>Which produces a high confidence level, the resolution list <figref>270</figref> added.

0045Thus, the merchant IDs represent on the resolution list <figref>270</figref> Merchant ID phrases <figref>210</figref>That the same company as the Example Merchant ID Set <figref>210 (0)</figref> belong.

0046The merchant-resolution tool <figref>150</figref> stores the resolution list <figref>270</figref> for use by the analytical engine <figref>155</figref>, Again, the analysis engine<figref>155</figref> the complete collection of transaction records <figref>215</figref> analyze the company regardless of the various merchant IDs in the transaction records <figref>215</figref> the Company are included. For example, should, if applicable, different merchant IDs in a resolution list<figref>270</figref> Transaction records <figref>215</figref> associate with multiple merchant accounts of several franchisees of a franchise company, then the analysis engine <figref>155</figref> the transaction records <figref>215</figref> with the merchant IDs in the resolution list <figref>270</figref> Merge to the complete collection of transaction records <figref>215</figref> analyze the franchise company.

0047<figref>3</figref> is a flowchart of method steps for training the classifier <figref>255</figref>According to one embodiment of the invention. Although the method steps are in connection with the systems of<figref>1</figref>-<figref>2</figref> and <figref>5</figref> described, however, it is clear to those skilled that any system configuration that performs the process steps in any order, is within the scope of the invention.

0048As shown, starts process <figref>300</figref> at step <figref>305</figref>In which a dealer-resolution tool <figref>150</figref> a training data set <figref>260</figref> positive examples of pairs of merchant id-sets <figref>210</figref> generated which are linked to the same company. The merchant-resolution tool<figref>150</figref> adds borderline cases to the training data set <figref>210</figref> added. The borderline cases include pairs of merchant id-sets<figref>210</figref>Which do not match, however, have similar merchant id strings. The limiting cases can also pairs of merchant id-sets<figref>210</figref> include having similar Aggregierungswerte, but are of different companies, so that they are actually negative training examples.

0049At step <figref>310</figref> adds the merchant-resolution tool <figref>150</figref> randomly selected pairs of merchant id-sets <figref>210</figref> the training data set <figref>260</figref> added. The randomly selected pairs of merchant id-sets<figref>210</figref> should mainly contain negative training examples.

0050At step <figref>315</figref> extends the merchant-resolution tool <figref>150</figref> respective merchant id phrases <figref>210</figref> the training data set <figref>260</figref> the Aggregierer <figref>240</figref> continue to candidate aggregations <figref>242</figref> to create. When training the classifier<figref>255</figref> there is no example Merchant ID Set <figref>210 (0)</figref>, And therefore are all merchant id phrases <figref>210</figref> the training data set <figref>260</figref> as candidates Merchant ID phrases <figref>210 (1)</figref> to <figref>210 (M - 1)</figref> considered. A user can this candidate aggregations<figref>242</figref> check.

0051At step <figref>320</figref> a user selects pairs of merchant id-sets <figref>210</figref>Which should be linked to the same company, as a positive training examples.

0052At step <figref>325</figref> a user selects pairs of merchant id-sets <figref>210</figref>That different Companies are linked, as negative training examples. These negative training examples contain several difficult borderline cases. In addition, the training data set includes<figref>210</figref> predominantly random selections, so that the majority of the pairs of sets merchant id <figref>210</figref> the training data set <figref>260</figref> negative training examples.

0053At step <figref>330</figref> training, merchant resolution tool <figref>150</figref> the classifier <figref>255</figref> with the training data set <figref>260</figref>, As described, the classifier<figref>255</figref> a Random Forest learning algorithm.

0054After training the classifier <figref>255</figref> with the training data set <figref>260</figref> , the classifier <figref>255</figref> a pair of merchant id-sets <figref>210</figref> evaluate, to generate a confidence, z. B. a value between zero and one. The confidence value is the percentage of decision trees in the classifier<figref>255</figref> used Random Forest algorithm that determine that both merchant id phrases <figref>210</figref> should be linked in the pair with the same company. Therefore, the classifier<figref>255</figref> able to produce a confidence value that represents whether a pair of merchant id sets <figref>210</figref>That a sample merchant id set <figref>210 (0)</figref> and a candidate Merchant ID Set <figref>210 (1)</figref> should include, linked to the same company.

0055<figref>4</figref> is a flow diagram of process steps to link merchant ID with a company according to an embodiment of the invention. Although the method steps are in connection with the systems of<figref>1</figref>-<figref>2</figref> and <figref>5</figref> described, however, it is clear to those skilled that any system configuration that performs the process steps in any order, is within the scope of the invention.

0056As illustrated, the process begins <figref>400</figref> at step <figref>410</figref>Wherein the merchant-resolution tool <figref>150</figref> an Example Merchant ID as merchant id attribute for a sample merchant id set <figref>210 (0)</figref> receives. As described, a user selects the Example Merchant ID Set<figref>210 (0)</figref> as representative of the characteristics of the financial transaction records <figref>215</figref> of which are associated with a company, z. B. the franchisee, which best represents a given franchise company. Alternatively, the system can automatically an Example Merchant ID Set<figref>210 (0)</figref> Select based on custom criteria.

0057In one embodiment, the merchant-resolution tool presents <figref>150</figref> the user a sample selection tool. The sample selection tool provides support when selecting a sample merchant id, which is representative of a company that is to be dissolved. The sample selection tool can obtain a search string from the user to identify merchant IDs that should be possibly linked to the company. The sample selection tool can also use any subset of the company name as a search string. In addition, the sample selection tool, the merchant id phrases<figref>210</figref>Associated with the identified merchant IDs to the Aggregierer <figref>240</figref> pass. The Aggregierer<figref>240</figref> then calculates aggregations <figref>242</figref>That assist the user in selecting the Example Merchant ID.

0058At step <figref>420</figref> generates the merchant-resolution tool <figref>150</figref> Sample aggregations <figref>244</figref> for the selected sample merchant id set <figref>210 (0)</figref>, After the dealer-resolution tool<figref>150</figref> the Example Merchant ID Set <figref>210 (0)</figref> from the transaction database <figref>160</figref> has accessed, calculates the Aggregierer <figref>240</figref> the average transaction size, the standard deviation of the transaction size and the average amount that an individual has spent.

0059At step <figref>430</figref> generates the merchant-resolution tool <figref>150</figref> Candidate aggregations <figref>242</figref> for a candidate Merchant ID Set <figref>210 (1)</figref>, The merchant-resolution tool<figref>150</figref> identifies a merchant id set <figref>210 (1)</figref> to <figref>210 (M - 1)</figref>Which does not match the sample merchant id set <figref>210 (0)</figref> was compared, as candidates Merchant ID Set <figref>210 (1)</figref>, Once the identification is done, calls the merchant resolution tool<figref>150</figref> Candidate Merchant ID Set <figref>210 (1)</figref> from the transaction database <figref>160</figref> from, and extends the candidate Merchant ID Set <figref>210 (1)</figref> the Aggregierer <figref>240</figref> continue. The Aggregierer<figref>240</figref> generates the candidate aggregations <figref>242</figref>,

0060Aggregating and comparing each potential merchant id record <figref>210 (1)</figref> to <figref>210 (M - 1)</figref> can be very time consuming, and therefore reducing the number of comparisons is desirable. In one embodiment, the merchant-resolution tool<figref>242</figref> not every merchant id record <figref>210</figref> compared. The merchant-resolution tool<figref>242</figref> skips merchant id records <figref>210</figref>Not satisfying certain conditions. Assuming that a franchise company has only franchisee locations in the state of California and the transaction records<figref>215</figref> an attribute to address include, where the transaction is carried out, then would the merchant resolution tool <figref>242</figref> Such merchant id records <figref>210</figref> Skip that no transaction records <figref>215</figref> California included. In this case, the merchant-resolution tool decreases<figref>242</figref> the number of comparisons, namely the fact that such merchant id phrases <figref>210</figref>, Which are not of California, are skipped.

0061At step <figref>440</figref> determines the dealer-resolution tool <figref>150</figref>Whether the Example Merchant ID Set <figref>210 (0)</figref> and the candidate merchant id set <figref>210 (1)</figref> match and should therefore be associated with the same company. The Identitätsauflöser<figref>250</figref> reaches the sample aggregations <figref>244</figref> and the candidate aggregations <figref>242</figref> to the classifier <figref>255</figref> continue. As described, the classifier is generated<figref>255</figref> a confidence between zero and one, of the percentage of decision trees in the classifier <figref>255</figref> used Random Forest algorithm corresponds which determine that both merchant id phrases <figref>210</figref> should be linked in the pair with the same company. If the classifier<figref>255</figref> generates a confidence value below a threshold value, then the method <figref>400</figref> step <figref>460</figref> continued. However, if the confidence value is above the threshold, then the method<figref>400</figref> step <figref>450</figref> continued. Although a threshold of 0.70 for the confidence value has proven to be effective, however, the actual threshold value can be set according to the preference.

0062At step <figref>450</figref> stores the Identitätsauflöser <figref>250</figref> the merchant id attribute of the candidate merchant id-set <figref>201 (1)</figref> in a resolution list <figref>270</figref>,

0063In one embodiment, the merchant-resolution tool performs <figref>242</figref> the Example Merchant ID Set <figref>240 (0)</figref> and candidate Merchant ID Set <figref>240 (1)</figref> to a combined merchant id-sentence, of a new and larger sample merchant id set <figref>240 (0)</figref> becomes. Then re-generates the merchant-resolution tool<figref>242</figref> the sample aggregations <figref>244</figref> for the remaining comparisons. This procedure allows the new Example Merchant ID Set<figref>240 (0)</figref> the company better represent and dissolving the remaining candidates Merchant ID phrases <figref>240 (2)</figref> to <figref>240 (M - 1)</figref> improve.

0064At step <figref>460</figref> determines the dealer-resolution tool <figref>150</figref>Whether there merchant id phrases <figref>210</figref> in the transaction database <figref>160</figref> out there that have not been compared. If the merchant-resolution tool<figref>150</figref> determines that there is a further candidate Merchant ID Set <figref>240 (2)</figref> compare available, then the process goes <figref>400</figref> back to step <figref>430</figref>, Once any candidate Merchant ID phrases<figref>210</figref> remain longer, to be compared, linked the merchant resolution tool <figref>150</figref> Merchant ID phrases <figref>210</figref>That in the resolution list <figref>270</figref> are listed for the company.

0065At step <figref>470</figref> linked the merchant resolution tool <figref>150</figref> the Example Merchant ID Set <figref>210 (0)</figref> with candidates Merchant ID-sets <figref>210 (1)</figref> to <figref>210 (M - 1)</figref>That in the resolution list <figref>270</figref> are listed. As described, the dissolution of the merchant id phrases may involve a list of merchant id attributes is stored, which the analytical engine<figref>155</figref> can use to the transaction records <figref>215</figref> the company identified. Alternatively, the dealer-resolution tool<figref>150</figref> the transaction records <figref>215</figref> Merchant ID phrases <figref>210</figref> on the resolution list <figref>270</figref> link to the company, in that it in an attribute of the transaction records <figref>215</figref> the company name inscribed, so that the analytical engine <figref>155</figref> a query of the transaction database <figref>160</figref> by the company belonging to the transaction records <figref>215</figref> can perform.

0066<figref>5</figref> shows an exemplary server computer system <figref>130</figref>, On the dealer-resolution tool <figref>150</figref> runs according to an embodiment. As shown, the server computer system comprises<figref>130</figref> a central processing unit (CPU) <figref>550</figref>, A network interface <figref>570</figref>, A random access memory <figref>520</figref> and a memory device <figref>530</figref>, Each with a coupling device (bus) <figref>540</figref> are connected. The server computer system<figref>130</figref> can also be an I / O device interface <figref>560</figref> include the input / output devices <figref>580</figref> (Z. B. keyboard, display and mouse) to the computer system <figref>130</figref> connects. Further, in the context of this disclosure, the machine elements in the server computer system<figref>130</figref> shown si nd, a physical computer system corresponding to (z. B. a system in a data center) or can be a virtual machine instance that is running in a computing cloud.

0067The CPU <figref>550</figref> introduces reading and storing in memory <figref>520</figref> stored program instructions, as well as a storing and reading in memory <figref>520</figref> underlying application data. The bus<figref>540</figref> is used to program instructions and application data between the CPU <figref>550</figref>, The I / O device interface <figref>560</figref>, The storage device <figref>530</figref>, The network interface <figref>570</figref> and the memory <figref>520</figref> transferred to. It should be noted that there is included, the CPU<figref>550</figref> is representative of a single CPU, multiple CPUs, a single CPU that has multiple cores, a CPU with associated memory management unit, and the like. Generally included here is the RAM<figref>520</figref> representative of a random access memory (RAM). The memory device<figref>530</figref> may be a disk drive storage device. Although shown as a single unit, the storage device<figref>530</figref> be a combination of permanently installed and / or removable storage devices, such as built-disk drives, removable memory cards or an optical storage device, a grid-connected storage (NAS) or a storage area network (SAN).

0068The communications between the client <figref>120</figref> and the merchant-resolution tool <figref>150</figref> are over the network <figref>180</figref> via network interface <figref>570</figref> transfer.

0069As explanatory diagram includes the memory <figref>520</figref> dealer-resolution tool <figref>150</figref>, Sample aggregations <figref>244</figref>, Candidate aggregations <figref>242</figref> and resolution list <figref>270</figref>, The merchant-resolution tool<figref>150</figref> even includes a Aggregierer <figref>240</figref> and a classifier <figref>225</figref>, The memory device<figref>530</figref> includes a training data set <figref>533</figref>, Which the dealer-resolution tool <figref>150</figref> used to the classifier <figref>225</figref> to train.

0070The Aggregierer <figref>240</figref> generates the sample aggregations <figref>244</figref> and the candidate aggregations <figref>242</figref> from the transaction records <figref>215</figref>Resulting from the transaction database <figref>160</figref> were retrieved. The merchant-resolution tool<figref>150</figref> are database queries over the network <figref>180</figref> the transaction database <figref>160</figref> via network interface <figref>570</figref> out. Once the Aggregierer<figref>240</figref> the sample aggregations <figref>244</figref> and candidate aggregations <figref>242</figref> has produced, used, merchant resolution tool <figref>150</figref> the classifier <figref>225</figref>To determine whether the merchant id phrases <figref>240</figref> should be linked to a company.

0071Although here in memory <figref>520</figref> shown, it is also possible that the merchant-resolution tool <figref>150</figref>That sample aggregations <figref>244</figref>That candidate aggregations <figref>242</figref> and the resolution list <figref>270</figref> are stored in memory <figref>520</figref>, In the storage means <figref>530</figref>, Or shared between memory <figref>520</figref> and storage means <figref>530</figref>, Similarly, the training data set<figref>533</figref> in memory <figref>520</figref>, The storage device <figref>530</figref>, Or shared between memory <figref>520</figref> and storage means <figref>530</figref>be saved.

0072In some embodiments, the database repository <figref>160</figref> in the memory means <figref>530</figref> . are In such a case, the database queries and subsequent answers via the bus<figref>540</figref> transfer. As described, the client can<figref>120</figref> on the server computer system <figref>130</figref> are, in which case the client <figref>120</figref> also in memory <figref>520</figref> would be stored and the user input / output devices <figref>580</figref> would use to communicate with the client <figref>120</figref> via the I / O device interface <figref>560</figref> interact.

0073Although the above aims to Described embodiments of the invention, other and further embodiments of the invention may be devised without departing from the basic scope thereof. For example, aspects of the invention may be implemented in hardware or in software, or in a combination of hardware and software. An embodiment of the invention may be implemented as a program product for use with a computer system. The / program (s) of the program product define functions of the embodiments (including the methods described herein) and can be on a variety of computer readable storage media. Examples of computer-readable storage media include (i) non-writable storage media (eg. As read-only memory devices within a computer, CD-ROMs, which can be read by a CD-ROM drive, a flash memory, ROM memory blocks or any type of non-volatile solid-state semiconductor memory); and (ii) writable storage media (eg. as floppy disks in a disk drive or a hard disk drive or any type of random access solid state semiconductor memory) on which alterable information is stored.

0074The invention was described above with reference to specific embodiments. However, those skilled, it is clear that various modifications and changes may be made thereto, without departing from the broader spirit and scope of the invention, which is set forth in the appended claims. The foregoing description and drawings are, accordingly rather illustrative and non-limiting.

0075Therefore, the scope of the invention is determined by the following claims.