a data-driven fuzzy rule-based approach for studentacademic performance evaluation

A Data-Driven Fuzzy Rule-Based
Approach for Student
Academic Performance Evaluation
Ernest Wu
Paper Reading
Outline
 Basic concepts of academic performance
evaluation.
 Basic concepts of Fuzzy Rule-Based System
 demonstration
 Data Driven FRBS
 Subsethood-Based Rule Generation
Algorithm (SBA)
 Weighted Subsethood-based Rule
Generation Algorithm
衡量學業表現的原因
 可以對學生的表現有更多瞭解
 老師可以藉此給予學生幫助
 學生也可以因此克服弱點，或是成為進步的誘因
 藉由學業表現對學生作出相應的決策
 學生成績不好，重修或留級
 也能用來衡量老師的表現
 老師對學生表現的幫助有多少
衡量學業表現的方式
 Formative assessment
 著重在教授的”過程”
 日常的活動:平時表現、平時報告、小考
 Summative assessment
 最後總結的成績
 通常都會綜合以上兩者
 面對不同的衡量目標，採行不同的方式

 Formative assessment可以提供feedback
 更全面的瞭解
Assessment Components (method)
 Series of tests and quizzes
 Portfolios
 Formal written examinations
 Individual Assignments and Coursework
 Group work
 Observation
 Theses and publishable materials
 Posters and oral presentation
衡量學業表現的表示法
 Single letter-grade (A, B, C, D, E, F)
 Nominal score (1, 2, …, 10)
 Single numerical score (100 percent)
 Linguistic terms ("Pass" and "Fail“)
 GPA (0.00~4.00)
目前來說，使用數值資料表示法來作進一步統計計算比較普遍。
階層式的衡量法
 各種的assessment components，可以使用階
層的方式，將它們匯總起來。
新方法產生的原因
 確認傳統衡量方式是否有問題。
 傳統方式無法面對不確定的評分，然而老師打分數本
來就是大概，因此使用fuzzy concepts可以面對這種
狀況。
 傳統方式只能處理數值資料，對於自然語言或是含糊
的詞彙比較不能處理，若能使用自然語言將會讓衡量
更有彈性。
 傳統方式缺乏比較的資訊 (與他人或自己比較)

criterion-reference evaluation
Æ
norm-referenced evaluation (Z-Score)
使用兩者的Combination以得到更多的資訊
使用Fuzzy Rule-Based System (FRBS)
新增的衡量方式
Fuzzy Rule-Based System
 Fuzzy set Theory
 Fuzzy membership functions
 Fuzzy logical operators
 Fuzzy IF-THEN rules
Fuzzy set theory
 傳統set theory—everything is precise
 Fuzzy set theory
Fuzzy Membership Functions
 measure linguistic variable
 A linguistic variable is defined as a variable whose values
are words or sentences in a natural or synthetic language
 choosing or generating an appropriate fuzzy membership
function to represent a linguistic term is very important.
Fuzzy Logical Operators
 Traditional logical operators
 Complement Æ negation
 Intersection Æ conjunction
 Union Æ disjunction
Fuzzy Logical Operators(2)
 Fuzzy Negation:
 Fuzzy Conjunction (t-norm):
 Fuzzy Disjunction (t-conorm):
Min-Max operators have been used widely probably because of their simplicity.

Fuzzy IF-THEN Rules
 “IF x is A THEN y is B" where A and B are
fuzzy sets.
 fuzzy IF-THEN rules are production rules
whose antecedents, consequences or both
are fuzzy.
 linguistic fuzzy model
 Mamdani-type FRBS
 Takagi-Sugeno-Kang (TSK) type FRBS
Fuzzy IF-THEN Rules(2)
 條件規則：IF 溫度 is A, THEN 壓縮機 is B
 狀態：溫度 is 高
 動作：壓縮機 is 打開
 其中A、B為模糊集合。IF的部分稱為前件部，而
THEN的部分則稱為後件部。
Mamdani-type FRBS
Demonstration
Traditional decision tree- training set
NoStrongHighMildRain
YesWeakNormalHotOvercast
YesStrongHighMildOvercast
YesStrongNormalMildSunny
YesWeakNormalMildRain
YesWeakNormalCoolSunny
NoWeakHighMildSunny
YesStrongNormalCoolOvercast
NoStrongNormalCoolRain
YesWeakNormalCoolRain
YesWeakHighMildRain
YesWeakHighHotOvercast

NoStrongHighHotSunny
NoWeakHighHotSunny
play ballWindHumidityTemperatureOutlook
Traditional decision tree- generate tree
 Outlook = Overcast: Yes (4.0)
 Outlook = Sunny:
| Humidity = High: No (3.0)
| Humidity = Normal: Yes (2.0)
 Outlook = Rain:
| Wind = Weak: Yes (3.0)
| Wind = Strong: No (2.0)
Traditional decision tree- testing
 Outlook: Sunny, Humidity: Normal
 Decision:
 Yes CF = 1.00 [ 0.50 - 1.00 ]
Traditional decision tree- generate rules
 Rule 1:
 Outlook = Sunny ,Humidity = High
-> class No [63.0%]
 Rule 5:
 Outlook = Rain , Wind = Strong
-> class No [50.0%]
 Rule 3:
 Outlook = Overcast
-> class Yes [70.7%]
 Rule 2:
 Humidity = Normal
-> class Yes [66.2%]
 Rule 4:
 Outlook = Rain, Wind = Weak

-> class Yes [63.0%]
Data Driven FRBS with WSBA method
Training Dataset
Testing Set