Kaggle Competition

import pandas as pd import numpy as np import matplotlib.pyplot as plt import seaborn as sns from scipy import stats from scipy.stats import chi2_contingency import statsmodels.api as sm from statsmodels.formula.api import ols %matplotlib inline sns.set_style('white')

train=pd.read_csv('data/train.csv') test=pd.read_csv('data/test.csv')

for i in [train, test]: i['date'] = pd.to_datetime(i['date'], format = "%Y/%m/%d") # Gestion des valeurs manquantes catégorielles i['cheveux'] = i['cheveux'].fillna(i['cheveux'].mode()[0]) i['sexe'] = i['sexe'].fillna(i['sexe'].mode()[0]) i['diplome'] = i['diplome'].fillna(i['diplome'].mode()[0]) i['specialite'] = i['specialite'].fillna(i['specialite'].mode()[0]) i['dispo'] = i['dispo'].fillna(i['dispo'].mode()[0]) # Gestion des valeurs manquantes numériques i['age'] = i['age'].fillna(i['age'].mean()) i['exp'] = i['exp'].fillna(i['exp'].mean()) i['salaire'] = i['salaire'].fillna(i['salaire'].mean()) i['note'] = i['note'].fillna(i['note'].mean()) # Gestion des outliers i['exp'] = np.where(i['exp'] < 0, i['age'] - 20, i['exp']) i['age'] = np.where(i['age'] < 16, i['exp'] + 20, i['age']) i['AGE_BINNED'] = pd.cut(i['age'], bins = np.linspace(16, 74, num = 11)) i['EXP_BINNED'] = pd.cut(i['exp'], bins = np.linspace(0, 23, num = 7))

pd.crosstab(train.specialite, train.sexe)

g, p, dof, expctd = chi2_contingency(pd.crosstab(train.specialite, train.sexe)) g, p

sns.boxplot(x="cheveux", y="salaire", data=train)

fit = ols('salaire ~ cheveux', data = train).fit() fit.summary()

sm.stats.anova_lm(fit, typ=1)

sns.lmplot(x="exp", y="note", data=train )

ols("note ~ exp", train).fit().summary()