DAT490 - Assignment 2

# Several imports which are used throughout the assignment from sklearn import linear_model from sklearn.linear_model import LinearRegression from sklearn.linear_model import Ridge import matplotlib.pyplot as plt import numpy as np import pandas as pd import statsmodels.api as sm from statsmodels.formula.api import ols from sklearn.datasets import load_iris import seaborn as sns #seaborn is a package for nice-looking graphics from sklearn import metrics from sklearn.linear_model import LogisticRegression from sklearn.model_selection import train_test_split from sklearn.neighbors import KNeighborsClassifier from sklearn import datasets # Read csv file as a data frame df1 = pd.read_csv("houses_data.csv")

# Creating two arrays from the data frame, where x is the input and y is the output x = np.array(df1['Area']).reshape((-1, 1)) #Reshaping the x-array so it is two-dimensional y = np.array(df1['Price']) # Create a linear regression model model = LinearRegression().fit(x,y) # Plotting the regresion line xfit = np.linspace(0, 300, 1000) yfit = model.predict(xfit[:, np.newaxis]) # Plotting a scatter plot of the model plt.scatter(x, y) plt.plot(xfit, yfit) plt.show() # Printing the slope and intercept values from the model print(model.coef_) print(model.intercept_)

# Predicting the outcome (price) of three different values of x (living area in square meters), using the predict function on the model print(model.predict([[100]])) print(model.predict([[150]])) print(model.predict([[200]]))

# Drawing a residual plot from the data frame where x = Area and y = Price sns.residplot(x = "Area", y = "Price", data = df1) plt.show()

# Load the iris dataset from the sklearn into variable "data" data = load_iris() # Iris data is split into Training sets and Test sets, where the test size is set to 0.25 x_train, x_test, y_train, y_test = train_test_split(data.data, data.target, test_size=0.25, random_state=0) # One-vs-rest instance of the model logisticRegr = LogisticRegression(multi_class='ovr', solver='liblinear') # The model is trained based on the data logisticRegr.fit(x_train, y_train) # Predictions on the entire data set predictions = logisticRegr.predict(x_test) # Score is measuring how accurate our model performs, i.e. the correct predictions / total number of data points score = logisticRegr.score(x_test, y_test) # Creating and plotting a confusion model cm = metrics.confusion_matrix(y_test, predictions) plt.figure(figsize=(9,9)) ax = sns.heatmap(cm, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('\nPredicted label'); plt.title('Accuracy Score: {0}'.format(score), size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show();

# Could have been done using a loop instead of repeated code # Creating the KNN Classifier for K = 1, uniform-based knn1_u = KNeighborsClassifier(n_neighbors=1) # Creating the KNN Classifier for K = 1, distance-based knn1_d = KNeighborsClassifier(n_neighbors=1, weights = 'distance') # Using training sets created in Q2 to train the model knn1_u.fit(x_train, y_train) knn1_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred1_u = knn1_u.predict(x_test) y_pred1_d = knn1_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 1, uniform-based: ",metrics.accuracy_score(y_test, y_pred1_u)) print("Accuracy for K = 1, distance-based: ",metrics.accuracy_score(y_test, y_pred1_d)) # ----------------------------------------------------------- # Creating the KNN Classifier for K = 3, uniform-based knn3_u = KNeighborsClassifier(n_neighbors=3) # Creating the KNN Classifier for K = 3, distance-based knn3_d = KNeighborsClassifier(n_neighbors=3, weights = 'distance') # Using training sets created in Q2 to train the model knn3_u.fit(x_train, y_train) knn3_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred3_u = knn3_u.predict(x_test) y_pred3_d = knn3_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 3, uniform-based: ",metrics.accuracy_score(y_test, y_pred3_u)) print("Accuracy for K = 3, distance-based: ",metrics.accuracy_score(y_test, y_pred3_d)) # ----------------------------------------------------------- # Creating the KNN Classifier for K = 21, uniform-based knn21_u = KNeighborsClassifier(n_neighbors=21) # Creating the KNN Classifier for K = 1, distance-based knn21_d = KNeighborsClassifier(n_neighbors=21, weights = 'distance') # Using training sets created in Q2 to train the model knn21_u.fit(x_train, y_train) knn21_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred21_u = knn21_u.predict(x_test) y_pred21_d = knn21_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 21, uniform-based: ",metrics.accuracy_score(y_test, y_pred21_u)) print("Accuracy for K = 21, distance-based: ",metrics.accuracy_score(y_test, y_pred21_d)) # ----------------------------------------------------------- # Creating the KNN Classifier for K = 35, uniform-based knn35_u = KNeighborsClassifier(n_neighbors=35) # Creating the KNN Classifier for K = 35, distance-based knn35_d = KNeighborsClassifier(n_neighbors=35, weights = 'distance') # Using training sets created in Q2 to train the model knn35_u.fit(x_train, y_train) knn35_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred35_u = knn35_u.predict(x_test) y_pred35_d = knn35_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 35, uniform-based: ",metrics.accuracy_score(y_test, y_pred35_u)) print("Accuracy for K = 35, distance-based: ",metrics.accuracy_score(y_test, y_pred35_d)) # ----------------------------------------------------------- # Creating the KNN Classifier for K = 50, uniform-based knn50_u = KNeighborsClassifier(n_neighbors=50) # Creating the KNN Classifier for K = 50, distance-based knn50_d = KNeighborsClassifier(n_neighbors=50, weights = 'distance') # Using training sets created in Q2 to train the model knn50_u.fit(x_train, y_train) knn50_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred50_u = knn50_u.predict(x_test) y_pred50_d = knn50_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 50, uniform-based: ",metrics.accuracy_score(y_test, y_pred50_u)) print("Accuracy for K = 50, distance-based: ",metrics.accuracy_score(y_test, y_pred50_d)) # ----------------------------------------------------------- # Creating the KNN Classifier for K = 100, uniform-based knn100_u = KNeighborsClassifier(n_neighbors=100) # Creating the KNN Classifier for K = 100, distance-based knn100_d = KNeighborsClassifier(n_neighbors=100, weights = 'distance') # Using training sets created in Q2 to train the model knn100_u.fit(x_train, y_train) knn100_d.fit(x_train, y_train) # Predicting the response for test dataset y_pred100_u = knn100_u.predict(x_test) y_pred100_d = knn100_d.predict(x_test) # Printing the accuracy of the classifier print("Accuracy for K = 100, uniform-based: ",metrics.accuracy_score(y_test, y_pred100_u)) print("Accuracy for K = 100, distance-based: ",metrics.accuracy_score(y_test, y_pred100_d))

# Confusion matrix for classification model where K = 1 # Also represents all the distance-based models since they all had the same score cmknn1 = metrics.confusion_matrix(y_test, y_pred1_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn1, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 1', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show(); # ---------------------------------------------------------------------------------------------- # Confusion matrix for classification model where K = 3 cmknn3 = metrics.confusion_matrix(y_test, y_pred3_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn3, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 3', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show(); # ---------------------------------------------------------------------------------------------- # Confusion matrix for classification model where K = 21 cmknn21 = metrics.confusion_matrix(y_test, y_pred21_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn21, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 21', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show(); # ---------------------------------------------------------------------------------------------- # Confusion matrix for classification model where K = 35 cmknn35 = metrics.confusion_matrix(y_test, y_pred35_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn35, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 35', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show(); # ---------------------------------------------------------------------------------------------- # Confusion matrix for classification model where K = 50 cmknn50 = metrics.confusion_matrix(y_test, y_pred50_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn50, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 50', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show(); # ---------------------------------------------------------------------------------------------- # Confusion matrix for classification model where K = 100 cmknn100 = metrics.confusion_matrix(y_test, y_pred100_u) plt.figure(figsize=(9,9)) ax = sns.heatmap(cmknn100, annot=True, linewidths=.5, square = True, cmap = 'Reds_r'); plt.ylabel('Actual label'); plt.xlabel('Predicted label'); plt.title('K = 100', size = 15); ax.xaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) ax.yaxis.set_ticklabels([data.target_names[0],data.target_names[1], data.target_names[2]]) plt.savefig('iris_cm.png') plt.show();