Walmart_Retail-model.py

#!/usr/bin/env python
# coding: utf-8

# In[28]:


# import important libraries
import pandas as pd 
from matplotlib import dates
from datetime import datetime
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sb


# ## **Walmart** 
# walmart is an American multinational retail corporation that operates a chain of hypermarkets, discount department stores, and grocery stores from the United states.

# <div>
# <img src="attachment:Full%20list%20of%20154%20Walmart%20stores%20to%20close%20across%20U_S_.jpg" width="900"/>
# </div>

# ### Dataset Description
# 
# This is the historical data that covers sales from 2010-02-05 to 2012-11-01, in the file Walmart_Store_sales. Within this file you will find the following fields:
# 
# Store - the store number
# 
# Date - the week of sales
# 
# Weekly_Sales -  sales for the given store
# 
# Holiday_Flag - whether the week is a special holiday week 1 – Holiday week 0 – Non-holiday week
# 
# Temperature - Temperature on the day of sale
# 
# Fuel_Price - Cost of fuel in the region
# 
# CPI – Prevailing consumer price index
# 
# Unemployment - Prevailing unemployment rate

# In[2]:


# import the dataset
data = pd.read_csv('Walmart_Store_sales.csv')


# In[3]:


data.head()


# ## Understanding Dataset

# In[4]:


data.info()


# In[56]:


data.max() #Finds the Maximum value in each column 


# In[57]:


# checking for missing values
data.isnull().sum()


# ### Preparation Dataset

# In[32]:


# Convert date to datetime format and show dataset information
data['Date'] =  pd.to_datetime(data['Date'])
data.info()


# In[33]:


# Splitting Date and create new columns (Day, Month, and Year)
data["Day"]= pd.DatetimeIndex(data['Date']).day
data['Month'] = pd.DatetimeIndex(data['Date']).month
data['Year'] = pd.DatetimeIndex(data['Date']).year
data


# #### QUESTION 1 :- Which store has maximum sales in this dataset?

# In[6]:


data.loc[data['Weekly_Sales'] ==  data['Weekly_Sales'].max()]
# used to find the row meeting the specific condition, 
# Here we are checking in column Weekly_Sales which row or store in particular is having the maximum Weekly_Sales.


# From above Figure we see that Store 14 has the maximum weekly sales.

# #### QUESTION 2 :- Which store has maximum standard deviation i.e., the sales vary a lot. Also, find out the coefficient of variance (C0V)

# In[7]:


#Here i am grouping by store and finding the standard deviation and mean of each store.
maxstd=pd.DataFrame(data.groupby('Store').agg({'Weekly_Sales':['std','mean']}))
#Just resetting the index.
maxstd = maxstd.reset_index()
#Now we know that CoV is std/ mean we are doing this for each store.
maxstd['CoV'] =(maxstd[('Weekly_Sales','std')]/maxstd[('Weekly_Sales','mean')]) *100
#finding the store with maximum standard deviation.
maxstd.loc[maxstd[('Weekly_Sales','std')]==maxstd[('Weekly_Sales','std')].max()]


# From above Figure we can conclude that sales in store 14 vary a lot

# #### QUESTION 3 :- Which store/s has good quarterly growth rate in Q3’2012.

# In[8]:


#Converting the data type of date column to dateTime 
data['Date'] = pd.to_datetime(data['Date'])

#defining the start and end date of Q3 and Q2
Q3_date_from = pd.Timestamp(date(2012,7,1))
Q3_date_to = pd.Timestamp(date(2012,9,30))
Q2_date_from = pd.Timestamp(date(2012,4,1))
Q2_date_to = pd.Timestamp(date(2012,6,30))

#Collecting the data of Q3 and Q2 from original dataset.
Q2data=data[(data['Date'] > Q2_date_from) & (data['Date'] < Q2_date_to)]
Q3data=data[(data['Date'] > Q3_date_from) & (data['Date'] < Q3_date_to)]

#finding the sum weekly sales of each store in Q2
Q2 = pd.DataFrame(Q2data.groupby('Store')['Weekly_Sales'].sum())
Q2.reset_index(inplace=True)
Q2.rename(columns={'Weekly_Sales': 'Q2_Weekly_Sales'},inplace=True)

#finding the sum weekly sales of each store in Q2
Q3 = pd.DataFrame(Q3data.groupby('Store')['Weekly_Sales'].sum())
Q3.reset_index(inplace=True)
Q3.rename(columns={'Weekly_Sales': 'Q3_Weekly_Sales'},inplace=True)

#mergeing Q2 and Q3 data on Store as a common column
Q3_Growth= Q2.merge(Q3,how='inner',on='Store')


# In[9]:


#Calculating Growth rate of each Store and collecting it into a dataframe  
Q3_Growth['Growth_Rate'] =(Q3_Growth['Q3_Weekly_Sales'] - Q3_Growth['Q2_Weekly_Sales'])/Q3_Growth['Q2_Weekly_Sales']
Q3_Growth['Growth_Rate']=round(Q3_Growth['Growth_Rate'],2)
Q3_Growth.sort_values('Growth_Rate',ascending=False).head(1)


# In[10]:


Q3_Growth.sort_values('Growth_Rate',ascending=False).tail(1)


# From above tables we can infer that Q3 growth rate is in losses .
# the Store 16 has the least loss of 3% compared the other stores and store 14 has highest loss of 18%.

# #### QUESTION 4:- Some holidays have a negative impact on sales. Find out holidays which have higher sales than the mean sales in non-holiday season for all stores together.

# In[13]:


#finding the mean sales of non holiday and holiday 
data.groupby('Holiday_Flag')['Weekly_Sales'].mean()


# In[14]:


#marking the holiday dates 
Christmas1 = pd.Timestamp(date(2010,12,31) )
Christmas2 = pd.Timestamp(date(2011,12,30) )
Christmas3 = pd.Timestamp(date(2012,12,28) )
Christmas4 = pd.Timestamp(date(2013,12,27) )

Thanksgiving1=pd.Timestamp(date(2010,11,26) )
Thanksgiving2=pd.Timestamp(date(2011,11,25) )
Thanksgiving3=pd.Timestamp(date(2012,11,23) )
Thanksgiving4=pd.Timestamp(date(2013,11,29) )

LabourDay1=pd.Timestamp(date(2010,2,10) )
LabourDay2=pd.Timestamp(date(2011,2,9) )
LabourDay3=pd.Timestamp(date(2012,2,7) )
LabourDay4=pd.Timestamp(date(2013,2,6) )

SuperBowl1=pd.Timestamp(date(2010,9,12) )
SuperBowl2=pd.Timestamp(date(2011,9,11) )
SuperBowl3=pd.Timestamp(date(2012,9,10) )
SuperBowl4=pd.Timestamp(date(2013,9,8) )

#Calculating the mean sales during the holidays
Christmas_mean_sales=data[(data['Date'] == Christmas1) | (data['Date'] == Christmas2) | (data['Date'] == Christmas3) | (data['Date'] == Christmas4)]
Thanksgiving_mean_sales=data[(data['Date'] == Thanksgiving1) | (data['Date'] == Thanksgiving2) | (data['Date'] == Thanksgiving3) | (data['Date'] == Thanksgiving4)]
LabourDay_mean_sales=data[(data['Date'] == LabourDay1) | (data['Date'] == LabourDay2) | (data['Date'] == LabourDay3) | (data['Date'] == LabourDay4)]
SuperBowl_mean_sales=data[(data['Date'] == SuperBowl1) | (data['Date'] == SuperBowl2) | (data['Date'] == SuperBowl3) | (data['Date'] == SuperBowl4)]
#
list_of_mean_sales = {'Christmas_mean_sales' : round(Christmas_mean_sales['Weekly_Sales'].mean(),2),
'Thanksgiving_mean_sales': round(Thanksgiving_mean_sales['Weekly_Sales'].mean(),2),
'LabourDay_mean_sales' : round(LabourDay_mean_sales['Weekly_Sales'].mean(),2),
'SuperBowl_mean_sales':round(SuperBowl_mean_sales['Weekly_Sales'].mean(),2),
'Non holiday weekly sales' : data[data['Holiday_Flag'] == 0 ]['Weekly_Sales'].mean()}
list_of_mean_sales


# From above Figure we can infer that the mean sales of thanks giving is more than the non holiday weekly sales .

# #### QUESTION 5 :-Provide a Monthly,Quaterly and Semester view of sales in units and give insights.

# In[68]:


#Monthly sales 
monthly = data.groupby(pd.Grouper(key='Date', freq='1M')).sum()# groupby each 1 month
monthly=monthly.reset_index()
fig, ax = plt.subplots(figsize=(8,6))
X = monthly['Date']
Y = monthly['Weekly_Sales']
plt.plot(X,Y)
plt.title('Month Wise Sales')
plt.xlabel('Monthly')
plt.ylabel('Weekly_Sales')


# We can observe from the Monthly Sales Graph that highest sum of sales is recorded in between jan-2011 to march-2011.

# In[69]:


#Quaterly Sales 
Quaterly = data.groupby(pd.Grouper(key='Date', freq='3M')).sum()
Quaterly = Quaterly.reset_index()
fig, ax = plt.subplots(figsize=(8,6))
X = Quaterly['Date']
Y = Quaterly['Weekly_Sales']
plt.plot(X,Y)
plt.title('Quaterly Wise Sales')
plt.xlabel('Quaterly')
plt.ylabel('Weekly_Sales')


# We can observe from the Quarterly Sales Graph that higest sum of sales is recorded in Q1 of 2011 and 2012.

# In[70]:


#Semester Sales 
Semester = data.groupby(pd.Grouper(key='Date', freq='6M')).sum()
Semester = Semester.reset_index()
fig, ax = plt.subplots(figsize=(8,6))
X = Semester['Date']
Y = Semester['Weekly_Sales']
plt.plot(X,Y)
plt.title('Semester Wise Sales')
plt.xlabel('Semester')
plt.ylabel('Weekly_Sales')


# We can Observe from Semester Sales graph that at beginning of 1st sem of 2010 and 1st sem of 2013 sales are lowest.

# ### Build  prediction models to forecast demand

# In[72]:


# find outliers 
fig, axs = plt.subplots(4,figsize=(6,18))
X = data[['Temperature','Fuel_Price','CPI','Unemployment']]
for i,column in enumerate(X):
    sb.boxplot(data[column], ax=axs[i])


# In[20]:


# drop the outliers     
data_new = data[(data['Unemployment']<10) & (data['Unemployment']>4.5) & (data['Temperature']>10)]
data_new


# In[21]:


# check outliers 
fig, axs = plt.subplots(4,figsize=(6,18))
X = data_new[['Temperature','Fuel_Price','CPI','Unemployment']]
for i,column in enumerate(X):
    sb.boxplot(data_new[column], ax=axs[i])


# ### Bulding the model

# In[22]:


# Import sklearn 
from sklearn.ensemble import RandomForestRegressor
from sklearn.model_selection import train_test_split
from sklearn import metrics
from sklearn.linear_model import LinearRegression


# In[50]:


# Select features and target 
X = data_new[['Store','Fuel_Price','CPI','Unemployment']]
y = data_new['Weekly_Sales']

# Split data to train and test (0.80:0.20)
X_train, X_test, y_train, y_test = train_test_split(X,y,test_size=0.2)


# In[52]:


# Linear Regression model
print('Linear Regression:')
print()
reg = LinearRegression()
reg.fit(X_train, y_train)
y_pred = reg.predict(X_test)
print('Accuracy:',reg.score(X_train, y_train)*100)


print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))
print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))


sb.scatterplot(y_pred, y_test);


# In[54]:


# Random Forest Regressor
print('Random Forest Regressor:')
print()
rfr = RandomForestRegressor(n_estimators = 400,max_depth=15,n_jobs=5)        
rfr.fit(X_train,y_train)
y_pred=rfr.predict(X_test)
print('Accuracy:',rfr.score(X_test, y_test)*100)

print('Mean Absolute Error:', metrics.mean_absolute_error(y_test, y_pred))
print('Mean Squared Error:', metrics.mean_squared_error(y_test, y_pred))
print('Root Mean Squared Error:', np.sqrt(metrics.mean_squared_error(y_test, y_pred)))


sb.scatterplot(y_pred, y_test);


# In[ ]:





# In[ ]: