Abstract: For sequentially input data string, the outliner and the change point are detected through calculation of the outlier score and the change point score by combining a time-series model learning device to learn the generation mechanism of the read data series as the time-series statistic model, a score calculator to calculate the outlier score of each data based on the time-series model parameter and the input data, a moving average calculator to calculate the moving average of the outlier score, a time-series model learning device to learn the generation mechanism of the moving average series as the time-series statistic model and the above score calculator that further calculates the outlier score of the moving average based on the moving average of the outlier score and outputs the result as the change point score of the original data.

Abstract: The outlier detection device for detecting abnormal data in a data set includes an outlier rule preservation unit for holding a set of rules characterizing abnormal data, a filtering unit for determining whether each data of the data set is abnormal data or not based on the rules held in the outlier rule preservation unit, a degree of outlier calculation unit for calculating a degree of outlier indicative of a degree of being abnormal with respect to each data determined not to be abnormal data by the filtering unit, a sampling unit for sampling each data whose degree of outlier is calculated by the degree of outlier calculation unit based on the calculated degree of outlier to apply a label indicating whether data is abnormal data or not to the sampled data, and a supervised learning unit for generating a new rule characterizing abnormal data by supervised learning based on a set of the respective data to which the label is applied and adding the new rule to the set of rules held in the outlier rule preserva

Abstract: An object of the present invention is to analyze questionnaire replies automatically from questionnaire reply statements including free reply description in natural language collected through a network, by using a text classification engine, and to issue the analysis results of the claimant as the knowledge in the rule formats. Questionnaire respondents send questionnaire reply statements from respondent computers. Questionnaire reply statements are accumulated in a database through a network. A text classification engine reads out accumulated questionnaire reply statements from the database, and learns the rules for classifying the questionnaire reply statements, and issues them to the claimant.

Abstract: A document classification system and method classifies words into word clusters. Word clusters are arranged for categories of documents and a word falls into a word cluster at a probability. A linear combination model using a distribution of word clusters in a category and a distribution of words in the word cluster probabilistically indicates whether a document is in a particular category.