final_project_uma_namboothiripad (3).py

# -*- coding: utf-8 -*-
"""Final Project_Uma Namboothiripad.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1-8ctPDO-F1GYjztowygbfzJI2i5vZT9J

In my current department, we are facing high levels of employee turnover.  We have limited budgets and some employees so not believe there is opportunity for growth.  The recent turnover has caused stress among the existing employees.  Leadership is concerned about how to stop this loss of talent.  Despite what they think, a pizza party just won't fix this.
"""

from google.colab import drive
drive.mount('/content/drive')

import numpy as np 
import pandas as pd 
from matplotlib import pyplot as plt
import seaborn as sns
import lightgbm as lgb
from statsmodels.tsa.holtwinters import SimpleExpSmoothing
from sklearn.metrics import mean_absolute_error
from statsmodels.tsa.holtwinters import ExponentialSmoothing
from statsmodels.tsa.statespace.sarimax import SARIMAX
from statsmodels.tsa.arima_model import ARIMA
from statsmodels.tsa.seasonal import seasonal_decompose
import statsmodels.api as sm
import itertools
from sklearn.model_selection import train_test_split
pip install -U scikit-learn


import warnings
pd.set_option('display.max_columns', None)
pd.set_option('display.width', 500)
import warnings
warnings.simplefilter(action='ignore', category=FutureWarning)
from pandas_profiling import ProfileReport
from scipy.stats import norm
from pandas_profiling import ProfileReport

#import pickle
#pickle_out = open("retension.pkl","wb")
#pickle.dump(X_train, pickle_out)
#pickle_out.close()

!pip install pycaret

!pip install datasets

from datasets import load_dataset
import pandas as pd
import multiprocessing as mp
import numpy as np
import pandas as pd
import time
import concurrent.futures
import time

url="https://huggingface.co/datasets/namb0010/Turnover/raw/main/finalproject.csv"
df0 = pd.read_csv(url)

df0.head()

"""A report to give me an idea of what the data looks like.

Some additional information on some of the values in the data.  For education this is what the categories mean: 1 = Below College, 2 = College, 3 = Bachelor, 4 = Masters, 5 = Phd
"""

profile = ProfileReport(df0)
profile

"""drop the data that won't matter for attrition.  Everyone works the standard hours and the other information is just identifying"""

df0=df0.drop(['EmployeeCount', 'EmployeeID', 'StandardHours'], axis = 1)

"""**Feature Engineering of the data**[link text](https://)"""

df0.describe()

"""Looked to see if the data has any NA's.  """

df0.isna().sum()

"""Before I change any of the values, I am going to make a copy of the dataframe just to reduce any issues I will have 2 hours from now. """

df = df0.copy(deep=True)

"""The NAs was only 19 of the rows.  I am replacing it with the average which was 2.69"""

df=df.replace(np.nan, "2.69")

df.dtypes

df['NumCompaniesWorked']=df['NumCompaniesWorked'].astype(float)

df['TotalWorkingYears']=df['TotalWorkingYears'].astype(float)

"""Turn the categorical values into ordinal """

cat_cols = ['Attrition', 'BusinessTravel', 'Department', 'EducationField', 'Gender', 'JobRole', 'MaritalStatus']

from tqdm import tqdm

from sklearn.preprocessing import OrdinalEncoder

# In loop to minimize memory use
for i, v in tqdm(enumerate(cat_cols)):
    df[v] = OrdinalEncoder(dtype="int").fit_transform(df[[v]])

df.dtypes

df.describe()

"""Notes on the categorical data that was turned into ordinal, so I don't forget what they mean tomorrow. 
1. Attrition - 0: No, 1: Yes
2. Business Travel - 0: Non-Travel, 1: Travel Frequently, 2: Travel-Rarely
3. Department- 0: HR, 1: Student Services, 2: Recruitment
4. Gender - 0: Female, 1: Male
5. Marital Status- 0: Divorced, 1: Married, 2: Single
"""

df.head()

"""Break up the data into training and testing """

from sklearn.model_selection import train_test_split
train, test = train_test_split(df, test_size=0.2)

train.describe()

test.describe()

"""Overall looking at some of the data, some initial observations.

-most staff live close to work, farthest out is 30 miles
-some disparity in income, percent raise
-Looks like most people have gotten promotions in the last two years; but have people who haven't been promoted for 15
-some outliers with the current manager - some have worked together for a long time

Overall - nothing that makes me concerned about the quality of the data.  Most companies have a pay disparity from the highest earner to the lowest. With the variety of educational backgrounds, from no college to Phd, we can expect that reflected in salary in this data.  Therefore, I am not removing anything from this dataset.  

"""

train['DistanceFromHome'].hist(bins=25, figsize=(10,5))

train['MonthlyIncome'].hist(bins=25, figsize=(10,5))

train['PercentSalaryHike'].hist(bins=25, figsize=(10,5))

train['TotalWorkingYears'].hist(bins=25, figsize=(10,5))

train['YearsAtCompany'].hist(bins=25, figsize=(10,5))

train['YearsSinceLastPromotion'].hist(bins=25, figsize=(10,5))

train['YearsWithCurrManager'].hist(bins=25, figsize=(10,5))

"""Remember that this was categorical data before so this is 18 different job titles.  If you want to see the more categorical data, go to the profile report at the top. """

train['JobRole'].hist(bins=25, figsize=(10,5))

df.corr()

"""Some notes on what the data indicates about the variables and relationships
-Age and Attrition
- Marital status and Attrition
- Total Working Years and Attrition
-Years at Company and Attrition
-Years with Current Manager and Attrition

As expected with some of the other relationships.  Age and years at the company, Age and number of companies worked, Age and Years Since Last Promotion, age and Years with Current Manager, department with education role, number of companies worked vs last promotion, number of companies worked vs years with current manager, years at company vs years with current manager, years at company vs years since last promotion 
"""

dfCorr = df.corr()
filteredDf = dfCorr[((dfCorr >= .10) | (dfCorr <= -.10)) & (dfCorr !=1.000)]
plt.figure(figsize=(30,10))
sns.heatmap(filteredDf, annot=True, cmap="Reds")
plt.show()

import matplotlib.cm as cm

sns.countplot(x='MaritalStatus',hue='Attrition', data=df0)
plt.title("Attrition and Marital Status")
plt.autoscale(enable=True) 
plt.show()

plt.autoscale(enable=True) 
sns.countplot(x='Age',hue='Attrition', data=df0)
plt.title("Attrition and Age")
plt.show()

plt.autoscale(enable=True)  
sns.countplot(x='YearsAtCompany',hue='Attrition', data=df0)
plt.title("Attrition and Years working at company")
plt.show()

plt.autoscale(enable=True) 
sns.countplot(x='TotalWorkingYears',hue='Attrition', data=df0)
plt.title("Attrition and Total Working Years")
plt.show()

"""staff stay with they like their manager"""

plt.autoscale(enable=True) 
sns.countplot(x='YearsWithCurrManager',hue='Attrition', data=df0)
plt.title("Attrition and Do You like your boss")
plt.show()

"""I was surprised that the raise did not really impact it """

plt.autoscale(enable=True) 
sns.countplot(x='PercentSalaryHike',hue='Attrition', data=df)
plt.title("Attrition and Raise")
plt.show()

"""I was surprised that the monthly income did not impact this as much """

df.plot.scatter(x='MonthlyIncome', y= 'Attrition')

"""The correlations did not indicate it above but I biased with the data

Start dividing up the training data and testing data for model creation
"""

X = train.drop('Attrition',axis=1)
y = train['Attrition']
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

from sklearn.metrics import accuracy_score, log_loss
from sklearn.neighbors import KNeighborsClassifier
from sklearn.svm import SVC, LinearSVC, NuSVC
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import RandomForestClassifier, AdaBoostClassifier, GradientBoostingClassifier
from sklearn.naive_bayes import GaussianNB
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.discriminant_analysis import QuadraticDiscriminantAnalysis
from xgboost import XGBClassifier
import xgboost

classifiers = [
    KNeighborsClassifier(3),
    SVC(kernel="rbf", C=0.025, probability=True),
    #NuSVC(probability=True),
    DecisionTreeClassifier(),
    RandomForestClassifier(),
    XGBClassifier(),
    AdaBoostClassifier(),
    GradientBoostingClassifier(),
    GaussianNB(),
    LinearDiscriminantAnalysis(),
    QuadraticDiscriminantAnalysis()]

log_cols=["Attrition", "Accuracy", "Log Loss"]
log = pd.DataFrame(columns=log_cols)

for clf in classifiers:
    clf.fit(X_train, y_train)
    name = clf.__class__.__name__
    
    print("="*30)
    print(name)
    
    print('****Results****')
    train_predictions = clf.predict(X_test)
    acc = accuracy_score(y_test, train_predictions)
    print("Accuracy: {:.4%}".format(acc))
    
    train_predictions = clf.predict_proba(X_test)
    ll = log_loss(y_test, train_predictions)
    print("Log Loss: {}".format(ll))
    
    log_entry = pd.DataFrame([[name, acc*100, ll]], columns=log_cols)
    log = log.append(log_entry)
    
print("="*30)

from pycaret.classification import *
s = setup(train, target = 'Attrition', silent = True)

new_model = create_model('et')

tuned_new_model = tune_model(new_model)

plot_model(tuned_new_model)

plot_model(tuned_new_model, plot = 'feature')

tuned_new_model.get_params()

classifiers1 = [
    RandomForestClassifier()
    ]

log_cols=["Attrition", "Accuracy", "Log Loss"]
log = pd.DataFrame(columns=log_cols)
X_train= X_train.drop(columns=['TrainingTimesLastYear', 'Department', 'Gender', 'EducationField', 'DistanceFromHome', 'Education', 'JobRole', 'PercentSalaryHike', 'Department', 'BusinessTravel', 'Education', 'NumCompaniesWorked'])
X_test= X_test.drop(columns=['TrainingTimesLastYear', 'Department', 'Gender', 'EducationField', 'DistanceFromHome', 'Education', 'JobRole', 'PercentSalaryHike', 'Department', 'BusinessTravel', 'Education', 'NumCompaniesWorked'])

#print(X_train)
for clf in classifiers1:
    clf.fit(X_train, y_train)
    name = clf.__class__.__name__
    
    print("="*30)
    print(name)
    
    print('****Results****')
    train_predictions = clf.predict(X_test)
    acc = accuracy_score(y_test, train_predictions)
    print("Accuracy: {:.4%}".format(acc))
    
    train_predictions = clf.predict_proba(X_test)
    ll = log_loss(y_test, train_predictions)
    print("Log Loss: {}".format(ll))
    
    log_entry = pd.DataFrame([[name, acc*100, ll]], columns=log_cols)
    log = log.append(log_entry)
    
print("="*30)

#print(X_train)
#print(X_test)

df.drop(['BusinessTravel', 'Gender', 'EducationField', 'DistanceFromHome', 'Education', 'JobRole', 'PercentSalaryHike', 'Department', 'BusinessTravel', 'Education', 'NumCompaniesWorked', 'TrainingTimesLastYear'], axis = 1)

model = RandomForestClassifier()
df_new = {'Age': 25, 'JobLevel': 3, 'MaritalStatus': 1, 'MonthlyIncome': 13000,'TotalWorkingYears': 4, 'YearsAtCompany': 2, 'YearsSinceLastPromotion': 1,  'YearsWithCurrManager': 2 }
df = df.append(df_new, ignore_index = True)

model.fit(X_train, y_train)

model.fit(X_train, y_train)

"""Okay lets see if this will predict anything.  I will be # the code below.  """

#X_test3 = [[25, 3, 1, 13000, 4, 2, 1, 2]]

Y = model.predict(X_test3)
print(Y)

!pip freeze > requirements.txt