[Py3.5] Regression - Features, Lables, Training and Test w/ svm.SVR

/*******************************************************************************************************************
-- Title : [Py3.5] Regression - Features, Lables, Training and Test w/ svm.SVR
-- Reference : pythonprogramming.net
-- Key word : quandl sklearn scikit-learn linear_model linear regression svm support vector machine data set
                  pre-processing label train test training testing svm.svr support vector regression linearregression
                  kernel 선형 회귀 선형 회귀 모델 서포트 벡터 머신 서포트벡터머신 트레이닝 테스팅 to_csv
                  cross-validation cross validation shift
*******************************************************************************************************************/

■ Scripts

# -*- coding: utf-8 -*-
 
import pandas as pd
import quandl
import math
import numpy as np
from sklearn import preprocessing, cross_validation, svm
from sklearn.linear_model import LinearRegression
 
# ------------------------------
# -- Set Dataframe Option
# ------------------------------
pd.set_option('display.height', 1000)
pd.set_option('display.max_rows', 500)
pd.set_option('display.max_columns', 500)
pd.set_option('display.width', 1000)
 
 
# ------------------------------
# -- Get Dataset
# ------------------------------
# 시나리오 : 일자별 주식 자료를 통해 몇일 후 주가를 예측
df = quandl.get("WIKI/GOOGL")
 
print(df.head())
print("... get_dataset", "." * 100, "\n")
 
 
# ------------------------------
# -- Manipulate Data
# ------------------------------
df = df[['Adj. Open', 'Adj. High', 'Adj. Low', 'Adj. Close', 'Adj. Volume']]
 
print(df.head())
print(",,, pair_down_data", "," * 100, "\n")
 
# -- Add Calculated Column
df['HL_PCT'] = (df['Adj. High'] - df['Adj. Low']) / df['Adj. Close'] * 100.0
df['PCT_change'] = (df['Adj. Close'] - df['Adj. Open']) / df['Adj. Open'] * 100.0
 
df = df[['Adj. Close', 'HL_PCT', 'PCT_change', 'Adj. Volume']]
 
print(df.head())
print(",,, add_calculated_columns", "," * 100, "\n")
 
# -- write to .csv
df.to_csv("manipulated_wiki_google.csv")
 
# -- Add Label Column
forecast_col = 'Adj. Close'
df.fillna(value=-99999, inplace=True)  # replace NaN to -99999
forecast_out = int(math.ceil(0.01 * len(df)))  # len() : df row count, math.ceil : 가장 작은 정수 올림
 
print("len(df):", len(df))
print("0.01*len(df):", 0.01 * len(df))
print("math.ceil(0.01*len(df)):", math.ceil(0.01 * len(df)))
print("forecast_out:", forecast_out)
 
# -- shift : +) 다음으로 행 이동, -) 이전으로 행 이동
# -- 몇일 후의 값을 예측하기 위한 LABEL을 임으로 이동시킴
df['label'] = df[forecast_col].shift(-forecast_out)
 
print(df.head())
print(",,, row_shift_for_label", "," * 100, "\n")
 
 
# ------------------------------
# -- Make Training and Testing Set
# ------------------------------
 
# -- drop NaN rows
df.dropna(inplace=True)
 
X = np.array(df.drop(['label'], 1))  # label만 빼고 Train Array로
y = np.array(df['label'])            # Label만 Test Array로
 
print("X:", X)
print("y:", y)
print(";;; array(train:X, label:y)", ";" * 100, "\n")
 
# -- Pre-Processing
X = preprocessing.scale(X)  # 기계학습 기능을 (-1)~(+1) 범위로 설정
y = np.array(df['label'])
 
print("X:", X)
print("y:", y)
print(";;; pre-processed X,y", ";" * 100, "\n")
 
# -- 모델 검증(Cross-Validation)
X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.2)
 
print("X_train:", X_train[:5], "\n")
print("X_test:", X_test[:5], "\n")
print("y_train:", y_train[:20], "\n")
print("y_test:", y_test[:20])
print(";;; cross-validation", ";" * 100, "\n")
 
 
# ------------------------------
# -- Model Fitting
# ------------------------------
 
# --
# -- SVM.SVR(Support Vector Regression)
# --
clf = svm.SVR()
clf.fit(X_train, y_train)  # fitting
 
confidence = clf.score(X_test, y_test)
 
print(confidence)
print("^^^ try_svm.svr", "^" * 100, "\n")
 
# --
# -- Linear Regression
# --
 
# -- Linear Regression(일반 처리)
clf = LinearRegression()
clf.fit(X_train, y_train)  # fitting
 
confidence = clf.score(X_test, y_test)
 
print(confidence)
print("^^^ try_linear_regression", "^" * 100, "\n")
 
# -- Linear Regression(쓰레드 지정)
clf = LinearRegression(n_jobs=-1)
clf.fit(X_train, y_train)  # fitting
 
confidence = clf.score(X_test, y_test)
 
print(confidence)
print("^^^ try_linear_regression with thread", "^" * 100, "\n")
 
 
# ------------------------------
# -- Training(Using Kernel)
# ------------------------------
for k in ['linear', 'poly', 'rbf', 'sigmoid']:
    clf = svm.SVR(kernel=k)
    clf.fit(X_train, y_train)
    confidence = clf.score(X_test, y_test)
    print(k, ":", confidence)
 
print("*** Training w/ Kernel", "*" * 100, "\n")

■ Files

manipulated_wiki_google.csv