# 資料科學期末報告 ---- ## Preparing ### (1) import Libraries  --- ### (2)Read the Data in Pandas dataFrame  --- ## Data Preprocessing ### convert the one-hot encoding labels into class labels  --- ### Describe the features  --- ### Calculating the faults  --- ### Calculation plot  --- ### the correlation matrix  * Correlation coeff of "TypeOfSteel_A300" and "TypeOfSteel_A400" is -1 so we can drop one of them * * Correlation coeff of "X_minimum" and "X_maximum" is -1 so we can drop one of them * * Correlation coeff of "Y_minimum" and "Y_maximum" is -1 so we can drop one of them --- ### Split train/test data  --- ### Applying Random Forest  ---  ### Assessing feature importance with random forest   * SigmoidOfAreas has low importance--->we remove it  * 1.X_Minimum 2.X_Maximum 3.Y_Maximum 4.Y_Minimum 5.Pixels_Areas 6.X-Perimeter 7.Y-Perimeter--->these are the important factor ### Applying Random Forest after removing least important feature #### Accuracy Chart  #### Confusion Matrix  #### Classification result