数値データの解析モジュール

import pandas as pd

import matplotlib.pyplot as plt

import numpy as np

def basic_graph(x,y):

    plt.hist(x)

    plt.show()

    plt.scatter(x,y)

    plt.show()

    plt.boxplot(np.array(x))

    plt.show()

def reg_line(x,y):

    from sklearn import linear_model

    model = linear_model.LinearRegression()

    tokutyou = [[xd] for xd in x]

    model.fit(tokutyou, y)

    plt.scatter(x,y)

    plt.plot(tokutyou, model.predict(tokutyou))

if __name__ == '__main__':

    shiken=pd.read_csv("*******")

    print('記述統計量')

    print(shiken.describe())

    print(shiken["math"])

    x = shiken["math"]

    y = shiken["goukei"]

    print('基本グラフ')

    basic_graph(x,y)

    print('散布図と回帰直線')

    reg_line(x,y)

    plt.show()

import iss_dstat as ids

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import sklearn.model_selection

from sklearn import linear_model

model_L=linear_model.LinearRegression()

def train_test(x,y):

    xg,xh,yg,yh=sklearn.model_selection.train_test_split(x,y)

    plt.scatter(xg,yg)

    plt.scatter(xh,yh)

    plt.show()

def nl_reg_line(x,y):  

    xg,xh,yg,yh=sklearn.model_selection.train_test_split(x,y)

    model_L.fit(xg,yg)

    tokutyou=[[xd] for xd in range(0,100)]

    yosoku_L=model_L.predict(tokutyou)

    plt.plot(tokutyou,yosoku_L)

def nl_reg_value(t,x,y):  

    xg,xh,yg,yh=sklearn.model_selection.train_test_split(x,y)

    model_L.fit(xg,yg)

    p=model_L.predict(t)[0]

    c=model_L.coef_

    return p,c

if __name__ == '__main__':

    shiken=pd.read_csv("**********")

    x = shiken["math"]

    y = shiken["goukei"]

    print("学習用データと検証用データ（ランダムに変化）")

    train_test(x,y)

    train_test(x,y)

    print("目的変数の予測(回帰直線と機械学習)")

    ids.reg_line(x,y)

    x = shiken[["math"]]

    nl_reg_line(x,y)

    plt.show()

    print("クロスバリデーション法による目的変数の予測")

    plt.scatter(x,y)

    for k in range(0,10):

        nl_reg_line(x,y)

    plt.show()

    print("クロスバリデーション法による目的変数の予測値")

    kitaichi=[]

    kiyoritu=[]

    for k in range(0,100):    

        reg_value=nl_reg_value([[80]],x,y)

        kitaichi.append(reg_value[0])

        kiyoritu.append(reg_value[1])

    plt.hist(kitaichi)

    plt.show()

    plt.boxplot(np.array(kiyoritu))

    plt.show()

    x = shiken[["math","chem"]]

    kitaichi=[]

    kiyoritu=[]

    for k in range(0,100):

        reg_value=nl_reg_value([[80,80]],x,y)

        kitaichi.append(reg_value[0])

        kiyoritu.append(reg_value[1])

    plt.hist(kitaichi)

    plt.show()

    plt.boxplot(np.array(kiyoritu))

    plt.show()

import pandas as pd

import numpy as np

import matplotlib.pyplot as plt

import iss_ml as ml

import iss_dstat as ids

if __name__ == '__main__':

    dash=pd.read_csv("************")

    dash=dash[dash.columns[[3,5,7,9]]].replace("-",np.nan).dropna() #データの前処理

    print("統計ダッシュボートからダウンロードしたデータによる予測")

    x=dash[dash.columns[0]]

    y=dash[dash.columns[3]]

    ids.reg_line(x,y)

    x=dash[[dash.columns[0]]]

    ml.nl_reg_line(x,y)

    plt.show()

    x =dash[dash.columns[[0,1,2]]]

    kitaichi=[]

    kiyoritu=[]

    for k in range(0,100):

        reg_value=ml.nl_reg_value([[8000,12000,15000]],x,y)

        kitaichi.append(reg_value[0])

        kiyoritu.append(reg_value[1])

    plt.hist(kitaichi)

    plt.show()

    plt.boxplot(np.array(kiyoritu))

    plt.show()