International Business Machines Corporation

Hirofumi Matsuzawa

Masayuki Numao

Hiroshi Masuda

Fumihiko Kitayama

Madoka Yuriyama

Takeshi Fukuda

Tetsuya Nasukawa

Hisashi Kashima

Tomio Amano

Masahiro Ohkawa

Shohei Hido

Masaru Yamamoto

Akiko Murakami

Harunobu Kubo

Takehiko Kobayashi

Natsumi Kurashima

Ysuyoshi Ide

Rei Suginaka

A method, a computer program product, and a computer system for evaluation and modification of training data for classifiers. A computer system applies an each piece of training data of an each class to multiple classifiers. The computer system performs evaluation and validation of the training data for the each class and defines at least one representative class for the each piece of the training data of the each class if classification precision of the each piece of the training data is greater than a predetermined threshold. The computer system modifies the training data, based on results of the evaluation and the validation of the training data for the each class. The computer system performs evaluation and validation of the training data for representative classes. The computer system modifies the training data, based on results of the evaluation and the validation of the training data for the representative classes.

Validating training data of classifiers

Detection of defects in an online system includes, for example, receiving, by one or more processor, a plurality of search conditions regarding a search screen of the online system, generating, by the one or more processor, a plurality of test models having factors based on the plurality of search conditions and levels having one or more of the search conditions, generating, by the one or more processor, a plurality of test cases based on the generated plurality of test models having the factors based on the plurality of search conditions and the levels having one or more of the search conditions, and executing, by the one or more processor, the plurality of test cases on the online system to detect defects in the online system.

Defect detection using test cases generated from test models

Different virtual labels, for example, like +1 and &#x2212;1, are assigned to two data sets. A change analysis problem for the two data sets is reduced to a supervised learning problem by using the virtual labels. Specifically, a classifier such as logical regression, decision tree and SVM is prepared and is trained by use of a data set obtained by merging the two data sets assigned the virtual labels. A feature selection function of the resultant classifier is used to rank and output both every attribute contributing to classification and its contribution rate.

Change analysis

An apparatus is provided with base table storage sections that store base tables and delta tables for the base tables, a summary table storage section that stores a summary table for storing results of queries to a plurality of base tables and delta information about the summary table, delta data processing sections that insert delta data of the base tables into the delta tables, and a delta computation processing section that generates delta information about the summary table. The delta computation processing section is provided with a generation section that generates delta information about a specified base table on the basis of an update that has been performed for the base table, in a situation where a subsequent update of the specified base table is permitted; and a control section that performs control so that, when a different base table is specified, delta information is generated in a different transaction.

Apparatus, method and program for refreshing a summary table

A system for tracing a cause of a phenomenon occurring in products produced in a production process chain is provided. The system is provided with a storage unit for storing a virtual attribute in association with corresponding second products, a receiving unit for receiving information for specifying third products in which a phenomenon occurs, a correlation calculation unit for calculating a correlation coefficient between the third products specified by the information for specifying the third products and the group of the second products having a common virtual attribute, for every kind of second products used in the third products, and a cause identification unit for identifying the second products belonging to the kind of the second products for which a maximum correlation coefficient is calculated as the cause of the phenomenon.

System, method, program for assigning virtual attribute to product, and system, method, and program for tracing cause of phenomenon occurring in product

It is an object of the present invention to find out parts to be a highly possible cause of failure without searching all of part data of all of products. Dispersed parts data on a parts tree are sequentially accessed from a set of known failed products, and part attribute values each having a higher support in the faulty product are extracted. In this process, a subset of parts used in the faulty product is also obtained simultaneously. The part attribute values having higher supports and the subset of parts used in the faulty product are represented as a tree in which a parts type serves as a node. Next, an information gain of a rule that having the two part attribute values is a cause of failure is calculated on two part attribute values having higher supports on the tree of the parts type. This calculation is locally performed on a common parent part of two parts and parts having a certain information gain is outputted as a cause of failure. How to select these two part attributes is performed in such a way that part attributes located closer to each other on the tree are first evaluated, and first found part attributes are made a candidate of a cause of failure.

Method and system for obtaining a combination of faulty parts from a dispersed parts tree

A candidate synonym acquisition device acquires a set of candidate synonyms similar to an input word for each writer from data for each writer, and acquires a set of candidate synonyms similar to the input word from a collective data. A generated candidate synonym set is inputted to a candidate synonym determination device to evaluate the candidate synonyms of the collective data. In the evaluation, the status of 'absolute' is given to a word matching a word ranked first in the candidate synonyms for each writer and the status of 'negative' is given to words matching words ranked second and lower therein.

Candidate synonym support device for generating candidate synonyms that can handle abbreviations, mispellings, and the like

A text-mining system and method automatically extracts useful information from a large set of tree-structured data by generating successive sets of candidate tree-structured association patterns for comparison with the tree-structured data. The number of times is counted that each of the candidate association patterns matches with a tree in the set of tree-structured data in order to determine which of the candidate association patterns frequently matches with a tree in the data set. Each successive set of candidate association patterns is generated from the frequent association patterns determined from the previous set of candidate association patterns.

Method and system for detecting frequent association patterns

The two-dimensional coordinates of a vertex are extracted in each of a top view and front view and, if their X-coordinates are equal to each other, the combination of their Y-coordinate values is determined to be the two-dimensional coordinates of a candidate vertex in a side view. Then, candidate line segments for the side view are extracted from the line segments connecting two candidate vertices, excepting not only those line segments for which no corresponding line segment exists in the top and front views, but a so those line segments for which corresponding horizontal or vertical line segments exist in the top and front views, and which a e not horizontal or vertical in the side view.

Drawing candidate line segments extraction system, drawing candidate line segments extraction method, solid model synthesis system, and solid model synthesis method

To provide a method for performing a plurality of aggregations in parallel and at a high speed, in a computer system so constructed that each of a plurality of processors connected across a network can use a memory area for itself and a part of the database for itself that includes data categorized into one or a plurality of groups, a method comprising the steps of: (a) ensuring space for storing results of M aggregate queries of the N aggregate queries (M is an integer equal to or less than N) in the memory area for itself in each processor; (b) executing all of the M aggregate queries for the part of the database for itself in each processor; (c) transmitting the results of the M aggregate queries executed by each processor to another processor for counting up and calculating of a final result for counting up; and (d) repeating the steps (a) to (c) until execution of the N aggregate queries is completed by each processor.

Method for executing aggregate queries, and computer system

A technique and system to enhance the synthesis of solid models from two drawings. The two-dimensional coordinates of a vertex are extracted in each of a top view and a front view and, if their x-coordinates are equal to each other, the combination of their Y-coordinate values is determined to be the two-dimensional coordinates of a candidate vertex in a side view. Then, candidate line segments for the side view are extracted from the line segments connecting two candidate vertices. The process does not extract those line segments for which no corresponding line segment exists in the top and front views and those line segments for which corresponding horizontal or vertical line segments exist in the top and front views, but which are not horizontal or vertical in the side view.

Drawing candidate line segments extraction system and method and solid

A technique to synthesize an accurate solid model from drawing data. Closed regions are detected based on top view data and a closed region list is generated. A tree structure is detected and the list is converted to symbol groups, each of which is a set of symbols representative of the elements of the contour line of the closed region and symbols representative of a connection relationship between the elements. When the converted symbol groups include a symbol group having a specific symbol, that symbol group is divided into a plurality of symbol groups based on that specific symbol. The symbol groups are matched, and the area of the closed region is detected. Symbol groups of closed regions of the same shape are classified and symbol groups of closed regions of the same area are classified. From the classified symbol groups, symbol groups of closed regions having a child of the closed region of the same shape and area are classified, and a solid model is constructed.

Growing community of inventors

Sagamihara, Japan

Hirofumi Matsuzawa