Calculation of Hit Ratio by Logistic Regreesion, RandomForest, SVM about samsung & hyundai

from __future__ import division

import os,sys,datetime
import numpy as np
import matplotlib.pyplot as plt
import matplotlib.dates as mdates
import pandas as pd
import pprint
import statsmodels.tsa.stattools as ts
from sklearn.ensemble import RandomForestClassifier
from sklearn.linear_model import LogisticRegression
from sklearn.lda import LDA
from sklearn.metrics import confusion_matrix
from sklearn.qda import QDA
from sklearn.svm import LinearSVC, SVC


def load_stock_data(file_name):
df = pd.read_pickle(file_name)
return df

#create dataset
def make_dataset(df, time_lags=1):
    df_lag = pd.DataFrame(index=df.index)
    df_lag["Close"] = df["Close"]
    df_lag["Volume"] = df["Volume"]

    df_lag["Close_Lag%s" % str(time_lags)] = df["Close"].shift(time_lags)
    df_lag["Close_Lag%s_Change" % str(time_lags)] = df_lag["Close_Lag%s" % str(time_lags)].pct_change()*100.0

    df_lag["Volume_Lag%s" % str(time_lags)] = df["Volume"].shift(time_lags)
    df_lag["Volume_Lag%s_Change" % str(time_lags)] = df_lag["Volume_Lag%s" % str(time_lags)].pct_change()*100.0

    df_lag["Close_Direction"] = np.sign(df_lag["Close_Lag%s_Change" % str(time_lags)])
    df_lag["Volume_Direction"] = np.sign(df_lag["Volume_Lag%s_Change" % str(time_lags)])

    return df_lag.dropna(how='any')

#split dataset
def split_dataset(df,input_column_array,output_column,spllit_ratio):
    split_date = get_date_by_percent(df.index[0],df.index[df.shape[0]-1],spllit_ratio)

    input_data = df[input_column_array]
    output_data = df[output_column]

    # Create training and test sets
    X_train = input_data[input_data.index < split_date]
    X_test = input_data[input_data.index >= split_date]
    Y_train = output_data[output_data.index < split_date]
    Y_test = output_data[output_data.index >= split_date]

    return X_train,X_test,Y_train,Y_test


def get_date_by_percent(start_date,end_date,percent):
    days = (end_date - start_date).days
    target_days = np.trunc(days * percent)
    target_date = start_date + datetime.timedelta(days=target_days)
    #print days, target_days,target_date
    return target_date


#forecasting models
def do_logistic_regression(x_train,y_train):
    classifier = LogisticRegression()
    classifier.fit(x_train, y_train)
    return classifier


def do_random_forest(x_train,y_train):
    classifier = RandomForestClassifier()
    classifier.fit(x_train, y_train)
    return classifier


def do_svm(x_train,y_train):
    classifier = SVC()
    classifier.fit(x_train, y_train)
    return classifier


#operate and evaluate the models

def test_predictor(classifier,x_test,y_test):
    pred = classifier.predict(x_test)

    hit_count = 0
    total_count = len(y_test)
    for index in range(total_count):
        if (pred[index]) == (y_test[index]):
            hit_count = hit_count + 1
 
    hit_ratio = hit_count/total_count
    score = classifier.score(x_test, y_test)
    #print "hit_count=%s, total=%s, hit_ratio = %s" % (hit_count,total_count,hit_ratio)

    return hit_ratio, score
    # Output the hit-rate and the confusion matrix for each model
 
    #print("%s\n" % confusion_matrix(pred, y_test))



if __name__ == "__main__":
    # Calculate and output the CADF test on the residuals

    avg_hit_ratio = 0  
    for time_lags in range(1,6):
        print("- Time Lags=%s" % (time_lags))

        for company in ['samsung','hyundai']:
            df_company = load_stock_data('%s_2010to2017.csv'%(company))

            df_dataset = make_dataset(df_company,time_lags)
            X_train,X_test,Y_train,Y_test = split_dataset(df_dataset,["Close_Lag%s"%(time_lags),"Volume_Lag%s"%(time_lags)],"Close_Direction",0.75)
            #print X_test

            lr_classifier = do_logistic_regression(X_train,Y_train)
            lr_hit_ratio, lr_score = test_predictor(lr_classifier,X_test,Y_test)

            rf_classifier = do_random_forest(X_train,Y_train)
            rf_hit_ratio, rf_score = test_predictor(rf_classifier,X_test,Y_test)

            svm_classifier = do_svm(X_train,Y_train)
            svm_hit_ratio, svm_score = test_predictor(rf_classifier,X_test,Y_test)

            print("%s : Hit Ratio - Logistic Regreesion=%0.2f, RandomForest=%0.2f, SVM=%0.2f" % (company,lr_hit_ratio,rf_hit_ratio,svm_hit_ratio))

- Time Lags=1

samsung : Hit Ratio - Logistic Regreesion=0.54, RandomForest=0.53, SVM=0.53

hyundai : Hit Ratio - Logistic Regreesion=0.47, RandomForest=0.47, SVM=0.47


- Time Lags=2
samsung : Hit Ratio - Logistic Regreesion=0.54, RandomForest=0.49, SVM=0.49
hyundai : Hit Ratio - Logistic Regreesion=0.47, RandomForest=0.47, SVM=0.47


- Time Lags=3
samsung : Hit Ratio - Logistic Regreesion=0.54, RandomForest=0.49, SVM=0.49
hyundai : Hit Ratio - Logistic Regreesion=0.46, RandomForest=0.45, SVM=0.45


- Time Lags=4
samsung : Hit Ratio - Logistic Regreesion=0.54, RandomForest=0.52, SVM=0.52
hyundai : Hit Ratio - Logistic Regreesion=0.47, RandomForest=0.45, SVM=0.45


- Time Lags=5
samsung : Hit Ratio - Logistic Regreesion=0.54, RandomForest=0.48, SVM=0.48
hyundai : Hit Ratio - Logistic Regreesion=0.47, RandomForest=0.40, SVM=0.40

Share this

ForecasterJ