Construct the Interval Rank Algorithm on the College Entrance Examination Strategies for 12th Grade Students

• Main Authors: Sheng-Yi Lin, 林勝義
• Other Authors: Yi-Ning Tu
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/89903391836490577529

碩士 === 輔仁大學 === 統計資訊學系應用統計碩士在職專班 === 103 === 12th Grade Students need to determine whether they have tested out their normal capability in the rest of the one semester. Moreover, they should decide whether they are going to prepare for the recommendation exam for school, personal application, or the advanced subjects test. The college entrance exam strategy is the most important thing that students care about. Generally speaking, teachers usually predict students’ levels by one or the long-term personal grade and rank in school. This research extends the concept of the K nearest neighbor method, and develops the Internal Rank Algorithms. This work uses within-group grades in school in the long term, and determine whether a student have tested out the normal capability according to the grade of general scholastic ability test within group. The Internal Rank Algorithms deals with the problem of determining students’ performance of general scholastic ability test within group by one grade. This research has found that using the grades of classmates’ stimulated exam to predict the students’ performance of general scholastic ability test is more accurate than using the grades of sectional exam. Its F-measure is 0.20671, which is larger than the expectation value 0.075. In addition, there is no significant difference by using classmates in the same class and interring classes to estimate p-value=0.122 of k-nearest neighbor.This is, it can infer that using the grade of stimulated exam to estimate k-nearest neighbor of interring classes. Next, using the rank difference of within-group and the whole to know that the pair is (x, y) = (0, 0) of the highest same proportion based on within-group grade and the whole grade. Among this, the accurracy which calculates by using the grade of stimulated exam as the the whole grade in school is larger than 80%. And then adjust within-group grade to revised within-group grade by using Internal Rank Algorithm. The revised within-group grade is the college entrance exam strategy which this research gives to students. The college entrance exam strategy is the result of the suggestions based on general scholastic ability test performance of k-nearest neighbor in the long term, students’ questionnaires as subjectivity, and students’ general scholastic ability test performance of that time as objectivity. This research is going to understand which grades can actually reflect the normal capability among these three grades. According to the experimental result, no matter in the training set or testing set, the proposed method is more than 85% among the 96 available samples. In other words, the proposed method is relatively better in the three grades.