DrinkingResgression.py

import csv
import os
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from sklearn.cross_validation import train_test_split
from sklearn.linear_model import LinearRegression
from sklearn.metrics import mean_squared_error
from sklearn.ensemble import RandomForestRegressor
from sklearn.tree import DecisionTreeRegressor
from sklearn.svm import SVR
from sklearn import neighbors
from sklearn.neural_network import MLPClassifier
from sklearn import tree
import pydotplus
import numpy as np


df = pd.read_csv("./metpath.csv")
#prints the shape of the dataset
print(df.shape)

# Print the first row of all the entries with 2 drinks.
# The .iloc method on dataframes allows us to index by position.
print(df[df["drinks"] < 5].iloc[0])
# Print the first row of all the entires with drinks greater than 2.
print(df[df["drinks"] > 5].iloc[0])
#prints a histogram of drinks recorded
plt.hist(df["drinks"])
plt.show()
print("MEAN of DRINKS: %d" % df["drinks"].mean())
df = df[df["drinks"] > 5]
df = df.dropna(axis=0)


#Generate a K-Cluster PCS graph to divide the data into unique clusters
kmeans_model = KMeans(n_clusters=5, random_state=1)
good_columns = df._get_numeric_data()
kmeans_model.fit(good_columns)
labels = kmeans_model.labels_
pca_2 = PCA(2)
plot_columns = pca_2.fit_transform(good_columns)
plt.scatter(x=plot_columns[:,0], y=plot_columns[:,1], c=labels)
plt.show()


##Print a list of correlaries to determine relationships
#between drinks/day and the metabolites
print(df.corr()["drinks"])
columns = df.columns.tolist()
columns = [c for c in columns if c not in ["drinks", "f"]]
target = "drinks"
#Split into train/test data. I opted to use a 80/20 train/test split to not overfit
train = df.sample(frac=0.8, random_state=1)
test = df.loc[~df.index.isin(train.index)]
print(train.shape)
print(test)
print(test.shape)



#Linear regression model did not give a very nice prediction
model = LinearRegression()
model.fit(train[columns], train[target])
print("$$$$$$$")
predictions = model.predict(test[columns])
print(predictions)
print(mean_squared_error(predictions, test[target]))
print(test[columns])
plt.scatter(train[target],train[columns]["mcv"],color="black")
plt.scatter(train[target],train[columns]["alkphos"],color="blue")
plt.scatter(train[target],train[columns]["sgpt"],color="red")
plt.scatter(train[target],train[columns]["sgot"],color="green")
plt.scatter(train[target],train[columns]["gammagt"],color="yellow")
plt.show()


#A random forrest implementation
forr = RandomForestRegressor(n_estimators=100, min_samples_leaf=10, random_state=1)
forr.fit(train[columns], train[target])
fpredictions = forr.predict(test[columns])
print(fpredictions)
print(mean_squared_error(fpredictions, test[target]))


#A decision tree regressor
dtree = DecisionTreeRegressor(max_depth=2)
dtree.fit(train[columns], train[target])
dpredictions=dtree.predict(test[columns])
print(dpredictions)
print(mean_squared_error(dpredictions, test[target]))
dot_data = tree.export_graphviz(dtree, out_file=None)
graph = pydotplus.graph_from_dot_data(dot_data)



####Used for tree plot
# graph.write_pdf("./outputs/DecisionTreeRegression.pdf")



#A SVR(rbf,linear, and polynomial kernels) implementation
svrr = SVR(kernel='rbf', C=1e3, gamma=0.1)
svrr.fit(train[columns], train[target])
rp = svrr.predict(test[columns])
print(rp)
print(mean_squared_error(rp, test[target]))




exit()