Sap Se

Sudhir Verma

Anviti Srivastava

Nirjar Gandhi

Yadesh Gupta

Martin Adam

Vitezslav Visek

Prabal Mahanta

Neha Garg

Sarma Adithe Venkata Ram

Jani Mahammad

Nitin Chavan

Ayushi Singla

Sidharth Gupta

Rohan Koul

Sumit Kumar

Vidur Sailendra Bhatnagar

Akash Gupta

Divyanshu Bajpai

Sven Herzberg

Pravesh Verma

Vidur Shailendra Bhatnagar

Shilpa Viswanadha

Systems, methods, and computer program products are provided for generation of keywords for a solution note for resolving an issue associated with a computing component. A dataset for training a keyword data model is received. The dataset includes a plurality of variables associated with one or more values. The keyword data model is configured for determination, as a function of one or more variables in the plurality of variables, of one or more keywords in a plurality of keywords associated with a computing solution in a plurality of computing solutions for resolving a problem with an operation of a computing component in a plurality of computing components. The keyword data model is trained using the received dataset and the keyword data model is applied to one or more variables in the received dataset to generate one or more keywords. One or more keywords associated with the computing solution is generated.

Intelligent keyword recommender

This disclosure describes implementations for categorizing and displaying a Gantt chart of categorized errors, as well as performing an odd pattern analysis to identify anomalous errors including obtaining a log file that includes a number of log entries where each log entry includes an error message. Converting each error message into an error vector using a predictive model. Determining a category for each error vector using a K-nearest neighbor algorithm, and generating a Gantt chart representing error messages arranged hierarchically by category as a function of time for display at a user interface.

Extended dynamic intelligent log analysis tool

Programs, systems, and methods for generating a support tree for automated resolution of user issues. In some embodiments, historical data may be obtained from a history of support interactions including historical issues and solutions to the historical issues. Most used solutions may be determined and stored with associated historical issues. Clusters of issues and solutions may be generated and labeled for generation of support tree nodes. Furthermore, solutions may be correlate such that a plurality of solutions may be provided for the resolving user issues. In some embodiments, issue data indicative of user issues may be provided by users. As the user provides issue data, deflections between nodes on the support tree may narrow a field of potential solutions to one or more solutions to be provided to the user.

Auto-generation of support trees

Techniques for implementing a dynamic intelligent log analysis tool are disclosed. In some embodiments, a computer system performs operations comprising: obtaining a log file comprising a plurality of log entries, each log entry comprising an error message; identifying a set of unique words from the error messages; for each error message, computing a term-frequency vector based on a frequency of occurrence for each unique word of the set of unique words in the error message; for each error message, computing a similarity measure between the term-frequency vectors of the error message and every other error message of the log entries; for each error message, computing a score based on a sum of the similarity measures; and displaying an indication of one or more of the error messages on a computing device based on the scores for the one or more of the error messages.

Dynamic intelligent log analysis tool

Systems, methods, and computer media are described for user risk assessment using similarity analysis. Records of transactions performed by a user while in previous enhanced application access sessions can be evaluated against records of transactions performed by other users in previous sessions. The more similar a user is to other users, the more likely it is the user was acting in a typical manner, and the less likely the user poses a security risk. A similarity analysis can be performed using a bipartite graph linking a group of users and a group of application transactions. By examining an edge between a user and a performed transaction, other edges (and corresponding other users) can be identified that also performed the transaction. A similarity score can be calculated based on the bipartite graph and can be used to determine a risk classification and allow or deny an enhanced application access session request.

Securing applications through similarity-based risk assessment

A computer system may obtain reference ticket data for reference tickets of an online system, with the reference ticket data indicating a reference issue for a reference component and a reference solution, and also obtain learning content items from a learning management system. The computer system may then create a mapping between the reference issues indicated by the reference ticket data for the reference tickets and the learning content items using an unsupervised machine learning algorithm. The computer system may detect target ticket data that has been provided by a user to the online system via a computing device, with the target ticket data indicating a target issue for a target component, identify a target solution for the target issue based on the target ticket data using the mapping, and cause a recommendation of the target solution to be displayed on the computing device.

Recommendations for information technology service management tickets

A business data database may store transaction business information of an enterprise along with existing structures, rules, and classification recommendations. A business rules framework agent platform may execute supervised machine learning and generate industry agnostic relationship scores and classification scores based on data in the business data database. The business rules framework agent platform may then optimize data and table structures, using relation graph-based evaluation, in accordance with taxonomy data and the classification scores. A rules and configuration database may store the optimized data and table structures from the business rules framework agent platform. Further, a business configuration and rules mapper platform may identify business configuration data and business rules classification data based on the optimized data and table structures. In some embodiments, the identified business configuration data and business rules classification data are used to automatically update the existing structures, rules, and classification recommendations in the business data database.

Automated master data classification and curation using machine learning

Methods and system are disclosed that relate to optimizing search for data. In one aspect, an attribute vector may include unique value identifiers and be associated with a dictionary structure. For a unique value identifiers stored in an attribute vector and associated with the dictionary structure, start address and end address associated with the unique value identifiers is computed. Based on the computation, a range of positional addresses associated with the unique value identifiers may be generated and stored in a data structure. Upon receiving a request to search for data, the range of positional addresses in which the unique value identifiers may be searched is determined. Based on the determination, a database search engine optimizes the search for data in the attribute vector.

Optimizing search for data

An attribute vector including value identifiers and corresponding to a dictionary structure is identified. A dictionary type encoding structure is generated by virtually partitioning the dictionary structure. The dictionary type encoding structure may include multiple dictionary types. Based on the dictionary encoding structure, the attribute vector may be split to generate multiple attribute vector blocks that may be identified by block transition indices. Based on the dictionary types in the dictionary encoding structure, the value identifiers in the attribute vector blocks are rearranged. Such a rearrangement optimizes the attribute vector for searching the value identifiers.

Virtual split dictionary for search optimization

Methods and system are disclosed that generate an execution schedule to optimize a transformation of business. In one aspect, from multiple tables residing in multiple databases and storing business data associate with multiple business management systems, dependencies between the tables may be determined based on attributes associated with the tables. When execution time for transforming business data exists, a decrease time algorithm or a critical path algorithm may be executed to generate execution schedule and to calculate processor idle times during the transformation of business data. Based on the calculated processor idle times, whether or not to execute a local optimization algorithm may be determined. Based on the determination, execution schedule that optimize the transformation of business data may be generated. The transformation of business data may be executed based to the generated execution schedule that optimizes a time consumed for transforming the business data in the tables.

Optimizing transformation of data

Methods and system are disclosed that compress a dataset and optimize the dataset for searching data. In one aspect, based on attributes associated with a unified dataset, the dataset including sensor node identifier data, sensor node timestamp data and sensor measurement data are determined. A suffix tree algorithm may be executed to determine data compression logic to compress the dataset including the sensor node identifier data and sensor node measurement data. A sensor node timestamp data compression model may be executed to compress the dataset associated with the sensor timestamp data. The compressed datasets including the sensor node identifier data, the sensor node timestamp data and the sensor measurement data may be stored in an in-memory data store.

Data compression

Methods and system are disclosed that optimize attribute vectors to search for data. In one aspect, a reordering optimization model is executed to reorder the attribute vectors. The attribute vectors may include value identifiers associated with data. The attribute vectors are assigned priorities, sequence score corresponding to each attribute vector is computed and the attribute vectors are reordered based on the computed sequence score. The reordered attribute vectors are logically partitioned into attribute vector blocks by executing a partitioning optimization model. The attribute vector blocks may be generated upon executing the partitioning optimization model. The value identifiers in the attribute vector blocks may be rearranged to optimize the attribute vectors to search for data.

Variable virtual split dictionary for search optimization

Various embodiments of systems and methods for optimizing in-memory column based database table compression. A system may receive a parsed in-memory columnar database table as an input and reads an aggregated compression score from memory corresponding to the parsed database table. A total compression score may be computed iteratively for the database table in order to determine whether the total compression score is greater than the aggregated compression score. When the total compression score determined to be greater than the aggregated compression score, the aggregated compression score is assigned to a best compression score. To optimize in-memory database table compression, the given database table may be compressed using the best compression score.

Optimizing column based database table compression

Database searching is optimized utilizing a graph community structure. A graph is created from transaction data based upon adjacent value occurrences. This may be done by scanning transaction data from top to bottom, and creating an edge between a current index value and a previous index value. Next, algorithms identify communities in the graph to create a graph community structure. These communities comprise blocks of patterns of similar value-ids distributed in the transaction data. Scanning and transition indices may be created with an eye toward reducing memory usage and enhancing performance. Query searching is then executed in an efficient manner on a per-community basis. For example, exact queries, range queries, and/or 'AND' queries may be executed more efficiently upon communities of records, skipping those not belonging to the same community. Embodiments are suited to search an in-memory database having large volumes of column-oriented data stored in RAM.

Growing community of inventors

Gurgaon, India

Sudhir Verma