eda_rf.py

"""
Final Project
EDA
"""


import pandas as pd
import matplotlib.pyplot as plt
from mlxtend.plotting import scatterplotmatrix
import numpy as np
import seaborn as sns
from imblearn.over_sampling import SMOTE
from sklearn.utils import resample
from mlxtend.plotting import heatmap
from sklearn.ensemble import RandomForestClassifier
from sklearn.preprocessing import StandardScaler, MinMaxScaler
from sklearn.feature_selection import SelectFromModel
import sys
from sklearn.model_selection import train_test_split
from collections import Counter

df = pd.read_csv('student-mat-edited.csv')


df['school'] = df['school'].replace(['GP', 'MS'], [1, 0])
df['sex'] = df['sex'].replace(['M', 'F'], [1, 0])
df['address'] = df['address'].replace(['U', 'R'], [1, 0])
df['famsize'] = df['famsize'].replace(['GT3', 'LE3'], [1, 0])
df['Pstatus'] = df['Pstatus'].replace(['T', 'A'], [1, 0])
df = df.replace(to_replace={'yes':1, 'no':0})

df = pd.get_dummies(df, prefix= ['Mjob', 'Fjob', 'reason', 'guardian'])

#code from: https://stackoverflow.com/questions/46168450/replace-a-specific-range-of-values-in-a-pandas-dataframe
#convert the scores to integers representing the letter grade range specified in the paper. higher the number, the higher the grade
df['scores'] = df[['G1', 'G2', 'G3']].mean(axis=1)
df['scores'] = np.where(df['scores'].between(0, 10), 0, df['scores'])
df['scores'] = np.where(df['scores'].between(10, 12), 1, df['scores'])
df['scores'] = np.where(df['scores'].between(12, 14), 2, df['scores'])
df['scores'] = np.where(df['scores'].between(14, 16), 3, df['scores'])
df['scores'] = np.where(df['scores'].between(16, 21), 4, df['scores'])
df['scores'] = df['scores'].astype(np.int)

df = df.drop(index=1, columns=['G1', 'G2', 'G3'])

#separate into features and target
X = df[[i for i in list(df.columns) if i != 'scores']]
y = df['scores']

# fixing class imbalance
#https://machinelearningmastery.com/multi-class-imbalanced-classification/
oversample = SMOTE(random_state=0)
X, y = oversample.fit_resample(X, y)

# splitting training and test data
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, random_state=0, stratify=y)


# min-max scaling
mms = MinMaxScaler()
X_train_norm = mms.fit_transform(X_train)
X_test_norm = mms.transform(X_test)

# standardizing the data
stdsc = StandardScaler()
X_train_std = stdsc.fit_transform(X_train)
X_test_std = stdsc.transform(X_test)


# Random Forest Feature Selection 

feat_labels = X.columns

forest = RandomForestClassifier(n_estimators=500, random_state=0) 
forest.fit(X_train, y_train)
importances = forest.feature_importances_
indices = np.argsort(importances)[::-1]
for f in range(X_train.shape[1]): 
    print("%2d) %-*s %f" % (f + 1, 30, feat_labels[indices[f]], importances[indices[f]])) 
plt.title('Feature Importance')
plt.bar(range(X_train.shape[1]), importances[indices], align='center')
plt.xticks(range(X_train.shape[1]), feat_labels[indices], rotation=90)
plt.xlim([-1, X_train.shape[1]]) 
plt.tight_layout()
plt.savefig("rf_selection.png")
plt.show()


sfm = SelectFromModel(forest, threshold=0.04, prefit=True)
X_selected = sfm.transform(X_train)
print('Number of features that meet this threshold', 'criterion:', X_selected.shape[1])



# # Now, let's print the  features that met the threshold criterion for feature selection that we set earlier (note that this code snippet does not appear in the actual book but was added to this notebook later for illustrative purposes):
cols = []
for f in range(X_selected.shape[1]):
     cols.append(feat_labels[indices[f]])    
     print("%2d) %-*s %f" % (f + 1, 30, 
                             feat_labels[indices[f]], 
                             importances[indices[f]]))


# Correlation heatmap
cols.append("scores")
cm = np.corrcoef(df[cols].values.T)
hm = heatmap(cm, row_names=cols, column_names=cols, figsize=(10, 8))
plt.savefig("corr_matrix.png")
plt.show()