Using the Text Mining Techniques to Construct the Personality Forecasting Model and Related Common Wordings Databases. In Cases of Lai’s Personality Test and Behavior-problem.

• Main Authors: Szu-Tzu Chung, 鍾思慈
• Other Authors: Yi-Ning Tu
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/37ze79

碩士 === 輔仁大學 === 統計資訊學系應用統計碩士在職專班 === 103 === In junior school age, the teacher plays an important role in the student’s growth process.At this stage, if the teacher can establish trust relationships with students, and timely support and care, they may have a chance to help them to develop a sound personality.If the teacher wanted to build a trust relationship with students, they have to observe mental state and personality traits of each student in a short time, and then use their own professional judgment counseling strategy. The TF-PDF (Term Frequency- Proportional Document Frequency) formula which was first introduced for news popularity analysis, and this work extended it as the new TF-PDF method for this study. This study hopes by using the TF-IDF and TF-PDF method of text mining techniques and the everyday diction of students to make a preliminary classification, since the diction may be hidden behind its mood and personality.This classification can help teachers to observe students' mental state and personality traits, and then determine whether students have long-term negative emotions which lead to behavior-problems. Besides, this study also constructs the personality forecasting model and related common wordings databases finally.When a student’s daily word is mapping to the vocabulary of mood and personality classification database, then this work can estimate the student should be assigned to which category. The results of this study found that, if use the TF-PDF method to construct the forecasting models, in "mood - psycho - social factors" the accuracy was 58.52%, while in the " behavior - problem leval " was 73.21%, both of them are better than the forecast model that uses TF-IDF method construction.Besides, no matter in what sort mode, using the TF-PDF method will get high TPR values that are more than 70%, but by TF-IDF method the TPR value was only 50%.The reason is that the latter will tend to classify the impact words into to the unrepresented vocabulary. Overview, the TF-PDF prediction model will have better predictive power.