kaggle研究生招生（上）

每天逛 kaggle

https://www.kaggle.com/mohansacharya/graduate-admissions

看来这个也是非常出名的数据集

GRE分数（290至340）
托福成绩（92-120）
大学评级（1至5）
目的声明（1至5）
推荐信强度（1至5）
本科生CGPA（6.8至9.92）
研究经验（0或1）
入学率（0.34至0.97）

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import seaborn as sns
import sys
import os

df = pd.read_csv("../input/Admission_Predict.csv",sep = ",")

硕士入学的三个最重要特征：CGPA、GRE和托福成绩

进入硕士学位的三个最不重要的特征：研究、LOR和SOP

相关系数矩阵

fig,ax = plt.subplots(figsize=(10, 10))
sns.heatmap(df.corr(), ax=ax, annot=True, linewidths=0.05, fmt= '.2f',cmap="magma")
plt.show()

但是数据大多数候选人都有研究经验。

因此，本研究将成为入学机会的一个不重要的特征

print("Not Having Research:",len(df[df.Research == 0]))
print("Having Research:",len(df[df.Research == 1]))
y = np.array([len(df[df.Research == 0]),len(df[df.Research == 1])])
x = ["Not Having Research","Having Research"]
plt.bar(x,y)
plt.title("Research Experience")
plt.xlabel("Canditates")
plt.ylabel("Frequency")
plt.show()

数据中托福最低分为92分，托福最高分为120分。平均107.41。

y = np.array([df["TOEFL Score"].min(),df["TOEFL Score"].mean(),df["TOEFL Score"].max()])
x = ["Worst","Average","Best"]
plt.bar(x,y)
plt.title("TOEFL Scores")
plt.xlabel("Level")
plt.ylabel("TOEFL Score")
plt.show()

GRE分数：

此柱状图显示GRE分数的频率。

密度介于310和330之间。在这个范围以上是候选人脱颖而出的一个很好的特征。

df["GRE Score"].plot(kind = 'hist',bins = 200,figsize = (6,6))
plt.title("GRE Scores")
plt.xlabel("GRE Score")
plt.ylabel("Frequency")
plt.show()

大学评分的CGPA分数：

随着大学质量的提高，CGPA分数也随之提高。

plt.scatter(df["University Rating"],df.CGPA)
plt.title("CGPA Scores for University Ratings")
plt.xlabel("University Rating")
plt.ylabel("CGPA")
plt.show()

GRE分数高的个体通常有较高的CGPA分数。

plt.scatter(df["GRE Score"],df.CGPA)
plt.title("CGPA for GRE Scores")
plt.xlabel("GRE Score")
plt.ylabel("CGPA")
plt.show()

df[df.CGPA >= 8.5].plot(kind='scatter', x='GRE Score', y='TOEFL Score',color="red")
plt.xlabel("GRE Score")
plt.ylabel("TOEFL SCORE")
plt.title("CGPA>=8.5")
plt.grid(True)
plt.show()

从好大学毕业的候选人更有幸被录取。

s = df[df["Chance of Admit"] >= 0.75]["University Rating"].value_counts().head(5)
plt.title("University Ratings of Candidates with an 75% acceptance chance")
s.plot(kind='bar',figsize=(20, 10))
plt.xlabel("University Rating")
plt.ylabel("Candidates")
plt.show()

CGPA分数高的候选人通常具有较高的SOP分数。

plt.scatter(df["CGPA"],df.SOP)
plt.xlabel("CGPA")
plt.ylabel("SOP")
plt.title("SOP for CGPA")
plt.show()

GRE分数高的候选人通常具有较高的SOP分数。

plt.scatter(df["GRE Score"],df["SOP"])
plt.xlabel("GRE Score")
plt.ylabel("SOP")
plt.title("SOP for GRE Score")
plt.show()

上面是数据分析过程，下面开始model的训练

去掉第一列的序号

# reading the dataset
df = pd.read_csv("../input/Admission_Predict.csv",sep = ",")# it may be needed in the future.
serialNo = df["Serial No."].valuesdf.drop(["Serial No."],axis=1,inplace = True)

y = df["Chance of Admit"].values
x = df.drop(["Chance of Admit"],axis=1)# separating train (80%) and test (%20) sets
from sklearn.model_selection import train_test_splitx_train, x_test,y_train, y_test = train_test_split(x,y,test_size = 0.20,random_state = 42)

缩放到固定范围（0-1）

# normalization
from sklearn.preprocessing import MinMaxScaler
scalerX = MinMaxScaler(feature_range=(0, 1))
x_train[x_train.columns] = scalerX.fit_transform(x_train[x_train.columns])
x_test[x_test.columns] = scalerX.transform(x_test[x_test.columns])

线性模型

from sklearn.linear_model import LinearRegression
lr = LinearRegression()
lr.fit(x_train,y_train)
y_head_lr = lr.predict(x_test)print("real value of y_test[1]: " + str(y_test[1]) + " -> the predict: " + str(lr.predict(x_test.iloc[[1],:])))
print("real value of y_test[2]: " + str(y_test[2]) + " -> the predict: " + str(lr.predict(x_test.iloc[[2],:])))from sklearn.metrics import r2_score
print("r_square score: ", r2_score(y_test,y_head_lr))y_head_lr_train = lr.predict(x_train)
print("r_square score (train dataset): ", r2_score(y_train,y_head_lr_train))

real value of y_test[1]: 0.68 -> the predict: [0.72368741]
real value of y_test[2]: 0.9 -> the predict: [0.93536809]
r_square score: 0.821208259148699
r_square score (train dataset): 0.7951946003191085

随机森林

from sklearn.ensemble import RandomForestRegressor
rfr = RandomForestRegressor(n_estimators = 100, random_state = 42)
rfr.fit(x_train,y_train)
y_head_rfr = rfr.predict(x_test) from sklearn.metrics import r2_score
print("r_square score: ", r2_score(y_test,y_head_rfr))
print("real value of y_test[1]: " + str(y_test[1]) + " -> the predict: " + str(rfr.predict(x_test.iloc[[1],:])))
print("real value of y_test[2]: " + str(y_test[2]) + " -> the predict: " + str(rfr.predict(x_test.iloc[[2],:])))y_head_rf_train = rfr.predict(x_train)
print("r_square score (train dataset): ", r2_score(y_train,y_head_rf_train))

r_square score: 0.8074111823415694
real value of y_test[1]: 0.68 -> the predict: [0.7249]
real value of y_test[2]: 0.9 -> the predict: [0.9407]
r_square score (train dataset): 0.9634880602889714

决策树

from sklearn.tree import DecisionTreeRegressor
dtr = DecisionTreeRegressor(random_state = 42)
dtr.fit(x_train,y_train)
y_head_dtr = dtr.predict(x_test) from sklearn.metrics import r2_score
print("r_square score: ", r2_score(y_test,y_head_dtr))
print("real value of y_test[1]: " + str(y_test[1]) + " -> the predict: " + str(dtr.predict(x_test.iloc[[1],:])))
print("real value of y_test[2]: " + str(y_test[2]) + " -> the predict: " + str(dtr.predict(x_test.iloc[[2],:])))y_head_dtr_train = dtr.predict(x_train)
print("r_square score (train dataset): ", r2_score(y_train,y_head_dtr_train))

r_square score: 0.6262105228127393
real value of y_test[1]: 0.68 -> the predict: [0.73]
real value of y_test[2]: 0.9 -> the predict: [0.94]
r_square score (train dataset): 1.0

线性回归和随机森林回归算法优于决策树回归算法。

y = np.array([r2_score(y_test,y_head_lr),r2_score(y_test,y_head_rfr),r2_score(y_test,y_head_dtr)])
x = ["LinearRegression","RandomForestReg.","DecisionTreeReg."]
plt.bar(x,y)
plt.title("Comparison of Regression Algorithms")
plt.xlabel("Regressor")
plt.ylabel("r2_score")
plt.show()

可视化三种算法

red = plt.scatter(np.arange(0,80,5),y_head_lr[0:80:5],color = "red")
green = plt.scatter(np.arange(0,80,5),y_head_rfr[0:80:5],color = "green")
blue = plt.scatter(np.arange(0,80,5),y_head_dtr[0:80:5],color = "blue")
black = plt.scatter(np.arange(0,80,5),y_test[0:80:5],color = "black")
plt.title("Comparison of Regression Algorithms")
plt.xlabel("Index of Candidate")
plt.ylabel("Chance of Admit")
plt.legend((red,green,blue,black),('LR', 'RFR', 'DTR', 'REAL'))
plt.show()