Group35_Assignment1

import pandas as pd from sklearn.model_selection import train_test_split # Read the CSV file. data = pd.read_csv('CTG.csv', skiprows=1) # Select the relevant numerical columns. selected_cols = ['LB', 'AC', 'FM', 'UC', 'DL', 'DS', 'DP', 'ASTV', 'MSTV', 'ALTV', 'MLTV', 'Width', 'Min', 'Max', 'Nmax', 'Nzeros', 'Mode', 'Mean', 'Median', 'Variance', 'Tendency', 'NSP'] #Filter missing data. data = data[selected_cols].dropna() # Shuffle the dataset. data_shuffled = data.sample(frac=1.0, random_state=0) # Split into input part X and output part Y. X = data_shuffled.drop('NSP', axis=1) # Map the diagnosis code to a human-readable label. def to_label(y): return [None, 'normal', 'suspect', 'pathologic'][(int(y))] Y = data_shuffled['NSP'].apply(to_label) # Partition the data into training and test sets. Xtrain, Xtest, Ytrain, Ytest = train_test_split(X, Y, test_size=0.2, random_state=0)

# We can also use call data_shuffled.head() to take a peek at the data set. X.head()

from sklearn.dummy import DummyClassifier clf = DummyClassifier(strategy='most_frequent')

from sklearn.model_selection import cross_val_score cross_val_score(clf, Xtrain, Ytrain, cv=5)

from sklearn.tree import DecisionTreeClassifier # No big difference if we set the value of max depth of 10. clf1 = DecisionTreeClassifier(max_depth=10) cross_val_score(clf1, Xtrain, Ytrain, cv=5)

from sklearn.ensemble import RandomForestClassifier clf2 = RandomForestClassifier(n_estimators = 100) cross_val_score(clf2, Xtrain, Ytrain, cv=5)

from sklearn.ensemble import GradientBoostingClassifier clf3 = GradientBoostingClassifier() cross_val_score(clf3, Xtrain, Ytrain, cv=5)

from sklearn.linear_model import Perceptron clf4 = Perceptron() cross_val_score(clf4, Xtrain, Ytrain, cv=5)

# Satndarlized data. from sklearn.linear_model import LogisticRegression from sklearn.preprocessing import StandardScaler Xtrain, Xtest, Ytrain, Ytest = train_test_split(StandardScaler().fit_transform(X), Y, test_size=0.2, random_state=0) clf5 = LogisticRegression(max_iter = 1000) cross_val_score(clf5, Xtrain, Ytrain, cv=5)

# Set max_iter to a larger valuefrom 1000 to 10000 and dual = True in LinearSVC classifier. from sklearn.svm import LinearSVC clf6 = LinearSVC(max_iter=10000, dual = True) cross_val_score(clf6, Xtrain, Ytrain, cv=5)

from sklearn.neural_network import MLPClassifier clf7 = MLPClassifier(max_iter=1000) cross_val_score(clf7, Xtrain, Ytrain, cv=5)

from sklearn.metrics import accuracy_score clf3.fit(Xtrain, Ytrain) Yguess = clf3.predict(Xtest) print(accuracy_score(Ytest, Yguess))

# Alternatively clf3.fit(Xtrain, Ytrain) clf3.score(Xtest, Ytest)

!pip install graphviz %run DecisionTree.py test = [] max_depth = 30 for i in range(max_depth): cls = TreeClassifier(max_depth=i+1) score = cross_val_score(cls, Xtrain, Ytrain, cv=5, scoring='accuracy').mean() test.append(score) max_depth_best = np.arange(1,max_depth+1)[np.argmax(test)]

import matplotlib.pyplot as plt plt.figure(figsize=(12,7)) plt.style.use('ggplot') plt.xticks(range(1, max_depth+1, 1)) plt.plot(range(1,max_depth+1),test,'bo-',label="max_depth") plt.xlabel('Max Depth', fontsize=15) plt.ylabel('Mean of 5-Fold Cross Validation Score', fontsize=15) plt.legend(fontsize=15, loc='lower right') plt.grid(True) plt.show() print("Best Cross Validation Score: {cv_score:.2f}% when max_depth = {md:d}"\ .format(cv_score=max(test)*100, md=max_depth_best))

cls = TreeClassifier(max_depth_best) cls.fit(Xtrain, Ytrain) Ypredict = cls.predict(Xtest) print("Best performance is {score:.2f}% when # of Tree = {treenum:d}"\ .format(score = accuracy_score(Ytest, Ypredict)*100, treenum = max_depth_best))

cls = TreeClassifier(np.floor(max_depth_best/4)) cls.fit(Xtrain, Ytrain) cls.draw_tree()

# Read the CSV file using Pandas. alldata = pd.read_csv('sberbank.csv') # Convert the timestamp string to an integer representing the year. def get_year(timestamp): return int(timestamp[:4]) alldata['year'] = alldata.timestamp.apply(get_year) # Select the 9 input columns and the output column. selected_columns = ['price_doc', 'year', 'full_sq', 'life_sq', 'floor', 'num_room', 'kitch_sq', 'full_all'] alldata = alldata[selected_columns] alldata = alldata.dropna() # Shuffle. alldata_shuffled = alldata.sample(frac=1.0, random_state=0) # Separate the input and output columns. X = alldata_shuffled.drop('price_doc', axis=1) # For the output, we'll use the log of the sales price. Y = alldata_shuffled['price_doc'].apply(np.log) # Split into training and test sets. Xtrain, Xtest, Ytrain, Ytest = train_test_split(X, Y, test_size=0.2, random_state=0)

from sklearn.dummy import DummyRegressor from sklearn.model_selection import cross_validate m1 = DummyRegressor() cross_validate(m1, Xtrain, Ytrain, scoring='neg_mean_squared_error')