Predictive Attrition Model: Using Analytics to predict Employee Attrition ¶

attrition = pd.read_csv('WA_Fn-UseC_-HR-Employee-Attrition.csv')

f, axes = plt.subplots(3, 3, figsize=(10, 8),
                       sharex=False, sharey=False)

# Defining our colormap scheme
s = np.linspace(0, 3, 10)
cmap = sns.cubehelix_palette(start=0.0, light=1, as_cmap=True)

# Generate and plot
x = attrition['Age'].values
y = attrition['TotalWorkingYears'].values
sns.kdeplot(x, y, cmap=cmap, shade=True, cut=5, ax=axes[0,0])
axes[0,0].set( title = 'Age against Total working years')

cmap = sns.cubehelix_palette(start=0.333333333333, light=1, as_cmap=True)
# Generate and plot
x = attrition['Age'].values
y = attrition['DailyRate'].values
sns.kdeplot(x, y, cmap=cmap, shade=True, ax=axes[0,1])
axes[0,1].set( title = 'Age against Daily Rate')

cmap = sns.cubehelix_palette(start=0.666666666667, light=1, as_cmap=True)
# Generate and plot
x = attrition['YearsInCurrentRole'].values
y = attrition['Age'].values
sns.kdeplot(x, y, cmap=cmap, shade=True, ax=axes[0,2])
axes[0,2].set( title = 'Years in role against Age')

cmap = sns.cubehelix_palette(start=1.0, light=1, as_cmap=True)
# Generate and plot
x = attrition['DailyRate'].values
y = attrition['DistanceFromHome'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[1,0])
axes[1,0].set( title = 'Daily Rate against DistancefromHome')

cmap = sns.cubehelix_palette(start=1.333333333333, light=1, as_cmap=True)
# Generate and plot
x = attrition['DailyRate'].values
y = attrition['JobSatisfaction'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[1,1])
axes[1,1].set( title = 'Daily Rate against Job satisfaction')

cmap = sns.cubehelix_palette(start=1.666666666667, light=1, as_cmap=True)
# Generate and plot
x = attrition['YearsAtCompany'].values
y = attrition['JobSatisfaction'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[1,2])
axes[1,2].set( title = 'Daily Rate against distance')

cmap = sns.cubehelix_palette(start=2.0, light=1, as_cmap=True)
# Generate and plot
x = attrition['YearsAtCompany'].values
y = attrition['DailyRate'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[2,0])
axes[2,0].set( title = 'Years at company against Daily Rate')

cmap = sns.cubehelix_palette(start=2.333333333333, light=1, as_cmap=True)
# Generate and plot
x = attrition['RelationshipSatisfaction'].values
y = attrition['YearsWithCurrManager'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[2,1])
axes[2,1].set( title = 'Relationship Satisfaction vs years with manager')

cmap = sns.cubehelix_palette(start=2.666666666667, light=1, as_cmap=True)
# Generate and plot
x = attrition['WorkLifeBalance'].values
y = attrition['JobSatisfaction'].values
sns.kdeplot(x, y, cmap=cmap, shade=True,  ax=axes[2,2])
axes[2,2].set( title = 'WorklifeBalance against Satisfaction')

f.tight_layout()

# Define a dictionary for the target mapping
target_map = {'Yes':1, 'No':0}
# Use the pandas apply method to numerically encode our attrition target variable
attrition["Attrition_numerical"] = attrition["Attrition"].apply(lambda x: target_map[x])

# creating a list of only numerical values
numerical = [u'Age', u'DailyRate', u'DistanceFromHome',
             u'Education', u'EmployeeNumber', u'EnvironmentSatisfaction',
             u'HourlyRate', u'JobInvolvement', u'JobLevel', u'JobSatisfaction',
             u'MonthlyIncome', u'MonthlyRate', u'NumCompaniesWorked',
             u'PercentSalaryHike', u'PerformanceRating', u'RelationshipSatisfaction',
             u'StockOptionLevel', u'TotalWorkingYears',
             u'TrainingTimesLastYear', u'WorkLifeBalance', u'YearsAtCompany',
             u'YearsInCurrentRole', u'YearsSinceLastPromotion',u'YearsWithCurrManager']
data = [
    go.Heatmap(
        z= attrition[numerical].astype(float).corr().values, # Generating the Pearson correlation
        x=attrition[numerical].columns.values,
        y=attrition[numerical].columns.values,
        colorscale='Viridis',
        reversescale = False,
#         text = True ,
        opacity = 1.0

    )
]


layout = go.Layout(
    title='Pearson Correlation of numerical features',
    xaxis = dict(ticks='', nticks=36),
    yaxis = dict(ticks='' ),
    width = 900, height = 700,

)


fig = go.Figure(data=data, layout=layout)
py.iplot(fig, filename='labelled-heatmap')

target_map = {'Yes':1, 'No':0}
# Use the pandas apply method to numerically encode our attrition target variable
target = attrition["Attrition"].apply(lambda x: target_map[x])

# from sklearn.cross_validation import train_test_split
from sklearn.model_selection import train_test_split
# Split data into train and test sets as well as for validation and testing
train, test, target_train, target_val = train_test_split(attrition_final,
                                                         target,
                                                         train_size= 0.80,
                                                         random_state=0);

oversampler=SMOTE(random_state=0)
smote_train, smote_target = oversampler.fit_resample(train,target_train)

seed = 0   # We set our random seed to zero for reproducibility
# Random Forest parameters
rf_params = {
    'n_jobs': -1,
    'n_estimators': 1000,
#     'warm_start': True, 
    'max_features': 0.3,
    'max_depth': 4,
    'min_samples_leaf': 2,
    'max_features' : 'sqrt',
    'random_state' : seed,
    'verbose': 0
}

rf = RandomForestClassifier(**rf_params)
rf.fit(smote_train, smote_target)
rf_predictions = rf.predict(test)
print("Accuracy score: {}".format(accuracy_score(target_val, rf_predictions)))
print("="*80)
print(classification_report(target_val, rf_predictions))