search

qyccc3 / CS334FinalProject

1 stars 0 forks source link

readme

CS334FinalProject House Sale Price Prediction

Instructions To Use Prediction App:

  1. git clone https://github.com/qyccc3/CS334FinalProject.git
  2. cd CS334FinalProject
  3. python HousePricePredictionApp.py

Make sure you have python3 installed and tkinter, pandas, sklearn library installed to use the application.

Data Processing and Models Used

Dataset: Ames Housing Dataset collecting from [https://www.kaggle.com/datasets/prevek18/ames-housing-dataset]()

We've manually selected and deleted the dataset from a total of 82 attributes to 31 attributes at the beginning.

import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
dataset = pd.read_csv("./ManualPreprocessedAmesHousing.csv")
dataset.head()
MS SubClass MS Zoning Lot Frontage Lot Area Lot Shape Lot Config Bldg Type House Style Overall Qual Overall Cond ... Paved Drive Wood Deck SF Open Porch SF Enclosed Porch 3Ssn Porch Screen Porch Pool Area Sale Type Sale Condition SalePrice
0 20 RL 141 31770 IR1 Corner 1Fam 1Story 6 5 ... P 210 62 0 0 0 0 WD Normal 215.0
1 20 RH 80 11622 Reg Inside 1Fam 1Story 5 6 ... Y 140 0 0 0 120 0 WD Normal 105.0
2 20 RL 81 14267 IR1 Corner 1Fam 1Story 6 6 ... Y 393 36 0 0 0 0 WD Normal 172.0
3 20 RL 93 11160 Reg Corner 1Fam 1Story 7 5 ... Y 0 0 0 0 0 0 WD Normal 244.0
4 60 RL 74 13830 IR1 Inside 1Fam 2Story 5 5 ... Y 212 34 0 0 0 0 WD Normal 189.9

5 rows × 55 columns

plt.figure(figsize=(30, 30))
sns.heatmap(dataset.corr())
plt.show()
/var/folders/kx/dd0tmbz95_n461n7xxtqy1wm0000gn/T/ipykernel_36849/2353200527.py:2: FutureWarning: The default value of numeric_only in DataFrame.corr is deprecated. In a future version, it will default to False. Select only valid columns or specify the value of numeric_only to silence this warning.
  sns.heatmap(dataset.corr())

png

dataset.hist(figsize=(30, 30))
plt.show()

png

for column in dataset.columns:
    if column in ['Gr Liv Area', '1st Flr SF', 'Garage Area', 'Overall Qual', 'Total Bsmt SF']:
        fig, ax = plt.subplots(figsize=(5,5))
        ax.scatter(dataset['SalePrice'], dataset[column])
        ax.plot
        ax.set_xlabel('SalePrice')
        ax.set_ylabel(column)
        plt.show()
# Only Show The Top 5 Correlated Features

png

png

png

png

png

After reviewing the scatter plots bewteen SalePrice and other columns, we picked the 5 most deteminant attributes that show high correlation with SalePrice: 'Gr Liv Area', '1st Flr SF', 'Garage Area', 'Overall Qual', 'Total Bsmt SF', 'SalePrice'

Models Used

Decision Tree\ Linear Regression\ KMeans

We categorized SalePrice for classification models based on 4 quantiles of the dataset:

house_data = pd.read_csv("./ManualPreprocessedAmesHousing.csv")[['Gr Liv Area', '1st Flr SF', 'Garage Area', 'Overall Qual', 'Total Bsmt SF','SalePrice']]
print("25 percentile of the data")
print('\t',int(house_data['SalePrice'].quantile(0.25)*1000), '== 0')
print("50 percentile of the data")
print('\t',int(house_data['SalePrice'].quantile(0.5)*1000), '== 1')
print("75 percentile of the data")
print('\t',int(house_data['SalePrice'].quantile(0.75)*1000), '== 2')
print("90 percentile of the data")
print('\t',int(house_data['SalePrice'].quantile(0.9)*1000), '== 3')
25 percentile of the data
     129500 == 0
50 percentile of the data
     160000 == 1
75 percentile of the data
     213500 == 2
90 percentile of the data
     281241 == 3
import pandas as pd
import numpy as np 
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from sklearn.model_selection import train_test_split
from sklearn.tree import DecisionTreeClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import mean_squared_error
from sklearn import tree
from sklearn.decomposition import PCA
from sklearn.linear_model import LinearRegression
from sklearn.model_selection import KFold
import graphviz
dataset = pd.read_csv("./ManualPreprocessedAmesHousingClassification.csv")[['Gr Liv Area', '1st Flr SF', 'Garage Area', 'Overall Qual', 'Total Bsmt SF','SalePrice']]

Decision Tree

from sklearn.model_selection import KFold
kf = KFold(n_splits=10, random_state=100, shuffle=True)
kf.get_n_splits(dataset)
depth = range(1,20)
accuracy = []
for max_d in depth:
    avg_accuracy = 0
    for train, test in kf.split(dataset):
        train_data = dataset.iloc[train]
        test_data = dataset.iloc[test]
        dt = DecisionTreeClassifier(criterion='gini', max_depth=max_d, random_state=100)
        dt.fit(train_data.drop('SalePrice', axis=1), train_data['SalePrice'])
        y_pred = dt.predict(test_data.drop('SalePrice', axis=1))
        # find average
        avg_accuracy += accuracy_score(test_data['SalePrice'], y_pred)
    accuracy.append(avg_accuracy/10*100)
plt.plot(depth, accuracy)
plt.xticks(depth)
plt.xlabel('Max Depth')
plt.ylabel('Accuracy')
plt.title('Accuracy vs Max Depth using Gini Index')
plt.show()

png

avg_accuracy = 0
print("10 Fold Cross Validation Using Gini Index, Max Depth 6: ")
for train, test in kf.split(dataset):
    train_data = dataset.iloc[train]
    test_data = dataset.iloc[test]
    dt = DecisionTreeClassifier(criterion='gini', max_depth=6, random_state=100)
    dt.fit(train_data.drop('SalePrice', axis=1), train_data['SalePrice'])
    y_pred = dt.predict(test_data.drop('SalePrice', axis=1))
    # find average
    avg_accuracy += accuracy_score(test_data['SalePrice'], y_pred)
    # print("\tAccuracy is ", accuracy_score(test_data['SalePrice'], y_pred)*100)
print("Average accuracy is ", avg_accuracy/10*100)
avg_accuracy = 0
print("10 Fold Cross Validation Using Entropy, Max Depth 6:")
for train, test in kf.split(dataset):
    train_data = dataset.iloc[train]
    test_data = dataset.iloc[test]
    dt = DecisionTreeClassifier(criterion='entropy', max_depth=6, random_state=100)
    dt.fit(train_data.drop('SalePrice', axis=1), train_data['SalePrice'])
    y_pred = dt.predict(test_data.drop('SalePrice', axis=1))
    # find average
    avg_accuracy += accuracy_score(test_data['SalePrice'], y_pred)
    # print("\tAccuracy is ", accuracy_score(test_data['SalePrice'], y_pred)*100)
print("Average accuracy is ", avg_accuracy/10*100)
10 Fold Cross Validation Using Gini Index, Max Depth 6: 
Average accuracy is  67.98634812286689
10 Fold Cross Validation Using Entropy, Max Depth 6:
Average accuracy is  66.9283276450512
# Best Max Depth is 6, using Gini Index
dt = DecisionTreeClassifier(criterion='gini', max_depth=6, random_state=100)
dt.fit(train_data.drop('SalePrice', axis=1), train_data['SalePrice'])
y_pred = dt.predict(test_data.drop('SalePrice', axis=1))
print("Accuracy is ", accuracy_score(test_data['SalePrice'], y_pred)*100)
Accuracy is  67.23549488054607
class_names = ['0', '1', '2', '3']
dot_data = tree.export_graphviz(dt, out_file=None,
                                feature_names = train_data.drop('SalePrice', axis=1).columns,
                                class_names = class_names,
                                filled = True, rounded = True,
                                special_characters = True)
graph = graphviz.Source(dot_data) 
graph

svg

KMeans

"""
Fit KMeans to the data without applying PCA
"""
kf = KFold(n_splits=10)
kf.get_n_splits(dataset)

avg_mse = 0
avg_acc = 0
for train, test in kf.split(dataset):
    train_data = dataset.iloc[train]
    test_data = dataset.iloc[test]
    kmeans = KMeans(n_clusters=4)
    kmeans.fit(train_data.drop("SalePrice", axis=1),train_data['SalePrice'])
    y_pred = kmeans.predict(test_data.drop("SalePrice", axis=1))
    # print("Mean Squared Error: ", mean_squared_error(y_pred, test_data['SalePrice']))
    # print("Arrucary: ", accuracy_score(y_pred, test_data['SalePrice']))
    avg_mse += mean_squared_error(y_pred, test_data['SalePrice'])
    avg_acc += accuracy_score(y_pred, test_data['SalePrice'])
print("Average Mean Squared Error: ", avg_mse/10)
print("Average Arrucary: ", avg_acc/10)
Average Mean Squared Error:  2.58122866894198
Average Arrucary:  0.22218430034129694
"""
Fit KMeans to the data after applying PCA
"""
pca = PCA(n_components=2)
pca.fit(dataset.drop('SalePrice', axis=1))
x_data = pca.transform(dataset.drop('SalePrice', axis=1))
x_data = pd.DataFrame(x_data)
y_data = dataset['SalePrice']
y_data = pd.DataFrame(y_data)
kf = KFold(n_splits=10)
kf.get_n_splits(dataset)
avg_mse = 0
avg_acc = 0
for train, test in kf.split(x_data):
    x_train = x_data.iloc[train]
    x_test = x_data.iloc[test]
    y_train = y_data.iloc[train]
    y_test = y_data.iloc[test]
    kmeans = KMeans(n_clusters=4)
    kmeans.fit(x_train,y_train)
    y_pred = kmeans.predict(x_test)
    # print("Mean Squared Error: ", mean_squared_error(y_pred, test_data['SalePrice']))
    # print("Arrucary: ", accuracy_score(y_pred, test_data['SalePrice']))
    avg_mse += mean_squared_error(y_pred, y_test)
    avg_acc += accuracy_score(y_pred, test_data['SalePrice'])
print("Average Mean Squared Error: ", avg_mse/10)
print("Average Arrucary: ", avg_acc/10)
Average Mean Squared Error:  2.1901023890784983
Average Arrucary:  0.2563139931740614

Linear Regression

# Linear Regression on Original Data
kFold = KFold(n_splits=10, shuffle=True, random_state=0)
dataset_orig = pd.get_dummies(pd.read_csv("./ManualPreprocessedAmesHousing.csv"))
X_orig = dataset_orig.drop(columns=["SalePrice"])
y_orig = dataset_orig["SalePrice"]
kFold = KFold(n_splits=10, shuffle=True, random_state=0)
avgMSE = 0
avgR2 = 0
for train, test in kFold.split(X_orig):
    X_train_fold, X_test_fold = X_orig.iloc[train], X_orig.iloc[test]
    y_train_fold, y_test_fold = y_orig.iloc[train], y_orig.iloc[test]
    model = LinearRegression()
    model.fit(X_train_fold, y_train_fold)
    y_pred = model.predict(X_test_fold)
    avgMSE += mean_squared_error(y_test_fold, y_pred)
    avgR2 += model.score(X_test_fold, y_test_fold)
print("Average MSE: ", avgMSE/10)
print("Average R2: ", avgR2/10)
Average MSE:  935.6220729544648
Average R2:  0.8555372323665559
"""
Linear Regression on Filtered Data
"""
dataset_filtered = pd.read_csv("./ManualPreprocessedAmesHousing.csv")[['Gr Liv Area', '1st Flr SF', 'Garage Area', 'Overall Qual', 'Total Bsmt SF','SalePrice']]
X_filtered = dataset_filtered.drop(columns=["SalePrice"])
y_filtered = dataset_filtered["SalePrice"]
kFold = KFold(n_splits=10, shuffle=True, random_state=0)
avgMSE = 0
avgR2 = 0
for train, test in kFold.split(X_filtered):
    X_train_fold, X_test_fold = X_filtered.iloc[train], X_filtered.iloc[test]
    y_train_fold, y_test_fold = y_filtered.iloc[train], y_filtered.iloc[test]
    model = LinearRegression()
    model.fit(X_train_fold, y_train_fold)
    y_pred = model.predict(X_test_fold)
    avgMSE += mean_squared_error(y_test_fold, y_pred)
    avgR2 += model.score(X_test_fold, y_test_fold)
print("Average MSE: ", avgMSE/10)
print("Average R2: ", avgR2/10)
Average MSE:  1428.9737943338919
Average R2:  0.7777471605626946
lr = LinearRegression()
lr.fit(dataset_filtered.drop('SalePrice', axis=1), dataset_filtered['SalePrice'])
y_pred = lr.predict(dataset_filtered.drop('SalePrice', axis=1))
print("Mean Squared Error: ", mean_squared_error(y_pred, dataset_filtered['SalePrice']))
print("R2: ", model.score(dataset_filtered.drop('SalePrice', axis=1), dataset_filtered['SalePrice']))
fig, ax = plt.subplots(figsize=(7, 7))
plt.scatter(dataset_filtered['SalePrice'], y_pred)
x = np.linspace(0, 600, 1000)
plt.plot(x, x, color='red')
plt.xlabel("Actual Sale Price")
plt.ylabel("Predicted Sale Price")
plt.show()
Mean Squared Error:  1400.2244916039772
R2:  0.7803537073215394

png