grade_predictor

# -*- coding: utf-8 -*-
"""Copy of Mini Proj 2.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1WfpMy-LZTesHU4cq5jLN9wCv2r8nCSy8

The two approcahes used are random forest and Naive Bayes, both of them can be found below.
"""

import matplotlib.pyplot as plt
import numpy as np
from sklearn import datasets, linear_model
from sklearn.metrics import mean_squared_error, r2_score
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
from sklearn.naive_bayes import GaussianNB
from sklearn import model_selection
from sklearn.model_selection import train_test_split
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
import seaborn as sns
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import classification_report
from sklearn.metrics import accuracy_score
from sklearn.model_selection import train_test_split
from matplotlib.colors import ListedColormap
from sklearn import neighbors, datasets
# Load the library with the iris dataset
from sklearn.datasets import load_iris

# Load scikit's random forest classifier library
from sklearn.ensemble import RandomForestClassifier

# Set random seed
np.random.seed(0)

df = pd.read_csv('/content/MP2_Data.csv')
print(df.head(2))

df.corr(method ='pearson')

df['Week_Stat0'] = df['Week1_Stat0']+df['Week2_Stat0']+df['Week3_Stat0']+df['Week4_Stat0']+df['Week5_Stat0']+df['Week6_Stat0']+df['Week7_Stat0']+df['Week8_Stat0']+df['Week9_Stat0']

df['Week_Stat0']

df['Week_Stat1'] = df['Week1_Stat1']+df['Week2_Stat1']+df['Week3_Stat1']+df['Week4_Stat1']+df['Week5_Stat1']+df['Week6_Stat1']+df['Week7_Stat1']+df['Week8_Stat1']+df['Week9_Stat1']

df['Week_Stat2'] = df['Week1_Stat2']+df['Week2_Stat2']+df['Week3_Stat2']+df['Week4_Stat2']+df['Week5_Stat2']+df['Week6_Stat2']+df['Week7_Stat2']+df['Week8_Stat2']+df['Week9_Stat2']

df['Week_Stat3'] = df['Week1_Stat3']+df['Week2_Stat3']+df['Week3_Stat3']+df['Week4_Stat3']+df['Week5_Stat3']+df['Week6_Stat3']+df['Week7_Stat3']+df['Week8_Stat3']+df['Week9_Stat3']

"""Feature selection"""

df1 = df[['Week8_Total','Week_Stat0','Week_Stat1','Week_Stat2','Week_Stat3','Grade']]
corela=df1.corr(method ='pearson')
round(corela,2)
corela

sns.heatmap(corela);

df1 = df1[['Week8_Total','Week_Stat0','Week_Stat1','Week_Stat3','Grade']]
#df1 = df[['Week_Stat0','Week_Stat3','Grade']]
#print(df1.head(5))
df1['is_train'] = np.random.uniform(0, 1, len(df1)) <= .75
print(df1.head(5))
train, test = df1[df1['is_train']==True], df1[df1['is_train']==False]

# Show the number of observations for the test and training dataframes
print('Number of observations in the training data:', len(train))
print('Number of observations in the test data:',len(test))

features = df1.columns[:4]
features

# Create a random forest Classifier. By convention, clf means 'Classifier'
clf = RandomForestClassifier(n_jobs=2, random_state=0)

# Train the Classifier to take the training features and learn how they relate
# to the training y (the species)
clf.fit(train[features], train['Grade'])

clf.predict(test[features])

clf.predict_proba(test[features])[0:10]

preds = clf.predict(test[features])
preds[0:10]

preds[0:10]

test['Grade'].head(30)

pd.crosstab(test['Grade'], preds, rownames=['Actual Grade'], colnames=['Predicted Grade'])

eg = pd.crosstab(test['Grade'], preds, rownames=['Actual Grade'], colnames=['Predicted Grade'])
sns.heatmap(eg)

length= len(preds)
length

preds
pred_clf_df = pd.DataFrame(preds.reshape(length,1))
pred_clf_df.rename(columns={0:'Prediction'}, inplace=True)
pred_clf_df

test[features]

#pred_outcome0 = pd.concat([test[features], pred_clf_df], axis=1)
#pred_outcome0 = pred_outcome0.reindex(test[features].index)
#pred_outcome0.rename(columns = {0:'Week2_Quiz1', 1:'Week4_Quiz2'}, inplace=True)
test['prediction'] = preds.tolist()
#test = test.drop(columns=['Grade'])
test

#plot the scatter plot of variable in data
test.plot.scatter(x="Grade",y="prediction")
plt.show()

#merging the prediction with original dataset
#pred_comp0 = pd.merge(df,pred_outcome0, on=['Week8_Total','Week_Stat0','Week_Stat3'])
pred_comp0 = pd.merge(df,test, on=['Week8_Total','Week_Stat0','Week_Stat1','Week_Stat3'])
#print top 10 lines of the final predictions
print((pred_comp0).head(10))
print ("\n")

#Save the file to csv
pred_comp0.to_csv('RF_Predictions.csv', sep=',')

accuracy_scorerf=accuracy_score(test['Grade'], preds)
accuracy_scorerf

list(zip(train[features], clf.feature_importances_))

# Set the figure size
plt.rcParams["figure.figsize"] = [7.00, 3.50]
plt.rcParams["figure.autolayout"] = True

# Plot bar chart with data points
plt.bar(features, clf.feature_importances_)

# Display the plot
plt.show()

"""Naive Bayes"""

df.head(2)

#Initialize Gaussian Naive Bayes
clf1 = GaussianNB()

clf1.fit(train[features], train['Grade'])

#Predicting for the Test Set
pred_clf1 = clf1.predict(test[features])


#Prediction Probability
prob_pos_clf1 = clf1.predict_proba(test[features])[:, 4]



pred_clf1

#prob_pos_clf1

pd.crosstab(test['Grade'], pred_clf1, rownames=['Actual Grade'], colnames=['Predicted Grade'])

htmpN = pd.crosstab(test['Grade'], pred_clf1, rownames=['Actual Grade'], colnames=['Predicted Grade'])
sns.heatmap(htmpN)

pred_clf1
length1= len(pred_clf1)
length1

#Create the prediction file by concatenation of the original data and predictions
#Reshaping needed to perform the concatenation
#pred_clf1_df = pd.DataFrame(pred_clf1.reshape(length1,1))

#Column renaming to indicate the predictions
#pred_clf1_df.rename(columns={0:'Prediction'}, inplace=True)
#pred_clf1_df

test[features]
test1=test
test1[features]

#concatenating the two pandas dataframes over the columns to create a prediction dataset
test1['prediction'] = pred_clf1.tolist()
#pred_outcome = pd.concat([test[features], pred_clf1_df], axis=1)
#pred_outcome = pred_outcome.reindex(test[features].index)
#pred_outcome.rename(columns = {0:'Week2_Quiz1', 1:'Week4_Quiz2'}, inplace=True)
test1

#plot the scatter plot of variable in data
test.plot.scatter(x="Grade",y="prediction")
plt.show()

#merging the prediction with original dataset
#pred_comp0 = pd.merge(df,pred_outcome0, on=['Week8_Total','Week_Stat0','Week_Stat3'])
pred_comp00 = pd.merge(df,test1, on=['Week8_Total','Week_Stat0','Week_Stat1','Week_Stat3'])
#print top 10 lines of the final predictions
print((pred_comp00).head(10))
print ("\n")

#Save the file to csv
pred_comp00.to_csv('NB_Predictions.csv', sep=',')

#Model Performance
#setting performance parameters
seed=7
kfold = model_selection.KFold(n_splits=10,shuffle=True, random_state=seed)

#calling the cross validation function
scoring='accuracy'
cv_results = model_selection.cross_val_score(GaussianNB(), train[features], train['Grade'], cv=kfold, scoring=scoring)

#displaying the mean and standard deviation of the prediction
msg = "%s: %f (%f)" % ('NB accuracy', cv_results.mean(), cv_results.std())
#msg1= " %f " % ( cv_results.mean())
print(msg)

list2=[]
list2.append(accuracy_scorerf)
msg1= " %f " % ( cv_results.mean())
msg1=float(msg1)
list2.append(msg1)
listname=["Random Forest","Naive Bayes"]

list2
#listname

# Set the figure size
plt.rcParams["figure.figsize"] = [4.00, 5.50]
plt.rcParams["figure.autolayout"] = True


# Plot bar chart with data points
plt.bar(listname, list2)
# giving X and Y labels

plt.ylabel("Accuracy Score")

# Display the plot
plt.show()