Fujitsu Corporation

Seishi Okamoto

Tatsuya Asai

Shinichiro Tago

Hiroya Inakoshi

Hiroaki Morikawa

Nobuhiro Yugami

Kanji Uchino

Tetsuro Takahashi

Takahisa Ando

Akira Sato

Tomoya Iwakura

Jun Zhang

Masahiko Nagata

Masataka Matsuura

Takeshi Osoekawa

Hiroya Hayashi

Satoko Shiga

Hiroki Arimura

A model updating method is provided. The model updating method that is executed by a computer includes calculating a score that indicates a degree of normality or abnormality of each of a plurality of pieces of data by using as a judgment model each of the pieces of data, predicting as a predicted condition whether each of the pieces of data is normal or abnormal according to score, judging whether or not the predicted condition is correct for each of the plurality of pieces of data, calculating the accuracy rate for the predicted conditions of a top specified number of pieces of data in order of decreasing abnormality as indicated by the score when the plurality of pieces of data are arranged in a specified order of score, and judging whether or not it is necessary to update the judgment model according to the accuracy rate.

Model updating method, model updating device, and recording medium

An observation information processing apparatus calculates, for each mesh, a support and a confidence. The observation information processing apparatus generates an adjacent mesh set by merging adjacent ones of the meshes. The observation information processing apparatus calculates, based on a support and a confidence of each mesh included in the adjacent mesh set, a confidence for each adjacent mesh, and sets the smallest one of the confidences calculated as a new confidence threshold value. The observation information processing apparatus detects and excludes meshes to be excluded from meshes included in the adjacent mesh set, based on the confidences and supports of the meshes included in the adjacent mesh set and the confidence threshold value.

Observation information processing apparatus, non-transitory computer-readable recording medium, and observation information processing method

A data partitioning apparatus receives stream data and stores with previous-and-subsequent information that specifies data previous and subsequent to the data. Then, for groups, each of which contains the data that are stored and that are partitioned into the number of the groups, the data partitioning apparatus stores top information specifying data located at the top in a corresponding group and end information specifying data located at the end in the corresponding group. Then, when new data is stored, the data partitioning apparatus specifies data previous and subsequent to the new data and stores the new data by associating it with the previous-and-subsequent information that specifies the data previous and subsequent to the new data. Then, when data is inserted, the data partitioning apparatus updates information for each group such that the difference between the number of data belonging to each group is equal to or less than one.

Data partitioning apparatus and data partitioning method

An apparatus includes a unit dividing a region into blocks having a predetermined bit length and generating addition data having the predetermined bit length corresponding to coordinate data, a unit storing storage data having the predetermined bit length, a unit generating composite data having a bit length twice the length of the predetermined bit length by connecting the addition data to the storage data, and storing the composite data, and a unit repeating an operation of designating the bit length of the composite data as the predetermined bit length, instructing to generate the addition data, and instructing to generate and store the composite data. The number of different bits between a bit string of a location corresponding to the input coordinate data and a bit string of a reference location tends to increase as a distance between the location corresponding to the coordinate data and the reference location increases.

Apparatus and method for coordinate coding, and method and apparatus for distance calculation

An apparatus extracts an area having a highest score or a lowest score that is calculated based on evaluation values of points from a target area including a set of a plurality of points that can be distributed at two-dimensional coordinates. It determines whether the area extracted and an area overlapping a plurality of the target areas intersect with each other, and deletes an area having a lower score or a higher score out of the areas determined to intersect with each other. It selects an area having a highest score or a lowest score out of the areas extracted at the extracting and not deleted at the determining. One or a plurality of areas are generated as the target areas based on an area acquired by excluding the area selected from the target area and an area acquired by excluding the area deleted at the determining.

Area finding apparatus, non-transitory computer readable storage medium, and method of finding area

A matching device builds a parse tree from an event pattern. The matching device builds an automaton by, while traversing the parse tree, joining child nodes of an AND node in parallel to the AND node and joining child nodes of an OR node in parallel to the OR node. The matching device performs matching of determining whether an event stream includes the event pattern by comparing the event stream with the automaton.

Matching device and computer-readable storage medium

A trajectory interpolation apparatus is disclosed. The first storage part stores first time and first location information of a movable body at the first time. The second storage stores second time and second location information of the movable body at the second time. The calculation part calculates a first moving distance from the first time and a second moving distance from the second time based on a relationship between the time and the speed stored in the second storage part, regarding third time between the first time and the second time. The determination part determines, as the interpolation point, one of intersection points for a circle in which the first location is set as its center and the first moving distance is set as its radius, and another circle in which the second location is set as its center and the second moving distance is set as its radius.

Trajectory interpolation apparatus and method

In a search device according to the present invention, a path trie generation unit generates an input path trie from XML data, a path trie display unit outputs the input path trie to an output unit such as a display, so that a user can specify Join points and combining conditions. An instruction reception unit generates a query serving as a command statement to combine each piece of XML data based on the Join points and the combining conditions that are specified by the user.

Computer readable storage medium recording database search program, database search device, and database search method

The matching apparatusstores, with respect to a branching pattern included in event patterns, a reception time of an event that serves as the branching point in a source node that indicates the transition destination of the event, connects the source node to the branch node by a transition with an interval condition, connects the branch nodes to the link node that indicates the linking point at which these branch nodes are linked together, and thereby generates NFAs. Then, the matching apparatuscompares event streamswith the NFAsto determine whether the event streamsincludes the event patterns

Matching pattern of events with interval conditions

When a searching apparatus accepts a search condition including a logical condition formed of variables, transition table is created by using a BDD method. In the transition table, a combination of a bit string indicative of true or false of a variable at each level for arriving from a starting-point variable to an arrival variable and the starting-point variable is associated with the arrival variable. Then, a true/false decision on the search condition of the search target data is made by, with an arrival variable of a transition destination being taken as the starting-point variable, extracting a bit string of a predetermined number of digits from the starting-point variable from among logical bit strings of the search target data and specifying the arrival variable as the transition destination based on the combination and the transition table.

True/false decision method for deciding whether search query containing logical expression is true or false

A retrieval device, from semi-structured data formed by item elements and value elements, creates aggregated structure information from the semi-structured data by aggregating the plurality of child item elements having the same name and directly connected to parent item elements having the same name, into one child item element. The device receives a retrieval request with an output condition and a constraint condition specified on the aggregate structure information. The device determines whether or not a plurality of query are created according to the retrieval request by checking a branch point at the common item elements between the output condition and the constraint condition in the aggregated structured information. And the device outputs a re-aggregate structure information corresponded to the query showing the branch point which is not expressed in the aggregate structure information to the display device.

Semi-structured data retrieval method, and structured date retrieval device

A rule learning method in machine learning includes distributing features to a given number of buckets based on a weight of the features which are correlated with a training example; specifying a feature with a maximum gain value as a rule based on a weight of the training example from each of the buckets; calculating a confidence value of the specified rule based on the weight of the training example; storing the specified rule and the confidence value in a rule data storage unit; updating the weights of the training examples based on the specified rule, the confidence value of the specified rule, data of the training example, and the weight of the training example; and repeating the distributing, the specifying, the calculating, the storing, and the updating, when the rule and the confidence value are to be further generated.

Rule learning method, program, and device selecting rule for updating weights based on confidence value

A keyword management system for managing keywords used when a user terminal connected to a network accesses contents, includes a Burst value calculating unit that calculates a Burst value indicating an increase per unit time of a keyword, an overall Burst value calculating unit that calculates an overall Burst value by correcting the Burst value based on characteristics in the contents of the keyword corresponding to the Burst value, and an output controlling unit that extracts from the contents, a relevant keyword related to the keyword corresponding to the overall Burst value based on time series changes of the overall Burst value, and outputs the keyword and the relevant keyword associated with each other to the user terminal.

Method and system for keyword management

A rule learning method for making a computer perform rule learning processing in machine learning includes firstly calculating an evaluation value of respective features in a training example by using data and weights of the training examples; selecting a given number of features in descending order of the evaluation values; secondly calculating a confidence value for one of the given number of selected features; updating the weights of training example, by using the data and weights of the training examples, and the confidence value corresponding to the one feature; firstly repeating the updating for the remaining features of the given number of features; and secondly repeating, for a given number of times, the firstly calculating, the selecting, the secondly calculating, the updating, and the firstly repeating.

Rule learning by updating training example and a remaining feature weights

Related keywords having certain relations to a target keyword are extracted from a unit storing keywords according to relations among each other. Expressions corresponding to the target keyword and related keywords are read from a unit storing expressions correspondingly with each stored keyword, each expression having been extracted from a content containing the corresponding keyword. A commonness between the expressions corresponding to each related keyword and the target keyword is calculated. An association between each related keyword and the target keyword is calculated using a distance for each related keyword from the target keyword and the commonness. A related keyword having a certain association is extracted as a same-region keyword in a same region as the target keyword. A drawing including the target keyword and related keywords arranged according to the associations is output with the same-region keyword displayed distinguishably from other related keywords not included in the same region.

Growing community of inventors

Kawasaki, Japan

Seishi Okamoto