Project Title coffee sales

Tools ML, SQL, Excel

Technologies Data Analyst & Data scientist

Project Difficulties level intermediate

Dataset : Dataset is available in the given link. You can download it at your convenience.

Click here to download data set

About Dataset
Overview

This dataset contains detailed records of coffee sales from a vending machine.
The vending machine is the work of a dataset author who is committed to providing an open dataset to the
community.
It is intended for analysis of purchasing patterns, sales trends, and customer preferences related to coffee products.

Data Collection Period

The dataset spans from March 2024 to Present time, capturing daily transaction data. And new information
continues to be added.

Tasks
 ● Time Series Exploratory Data Analysis
● Next day/week/month sales
● Specific customer purchases

Author

Yaroslav Isaienkov @ihelon

NOTE :
1. this project is only for your guidance, not exactly the same you have to create. Here I am trying to show the
way or idea of what steps you can follow and how your projects look. Some projects are very advanced (because it
will be made with the help of flask, nlp, advance ai, advance DL and some advanced things ) which you can not understand .
2. You can make or analyze your project with yourself, with your idea, make it more creative from where we can
get some information and understand about our business. make sure what overall things you have created all
things you understand very well.

Sure! Below is a step-by-step guide to performing a coffee sales analysis using

Python, focusing on data cleaning and basic machine learning (ML) modeling. This
example uses pandas for data manipulation and scikit-learn for machine learning. I'll
assume you have a dataset named coffee_sales.csv.

1. Data Collection
First, ensure you have the necessary libraries installed:

bash
Copy code
pip install pandas scikit-learn matplotlib seaborn

2. Data Preparation and Cleaning

Load and inspect the data:

import pandas as pd

# Load the dataset

data = pd.read_csv('coffee_sales.csv')

# Display the first few rows

print(data.head())

Assume the dataset has the following columns: Date, Store, Product, Sales,
Quantity, Price.

Handling Missing Values

# Check for missing values

print(data.isnull().sum())

# Fill missing numerical values with the median

data['Sales'].fillna(data['Sales'].median(), inplace=True)
data['Quantity'].fillna(data['Quantity'].median(),
inplace=True)
data['Price'].fillna(data['Price'].median(), inplace=True)

# Fill missing categorical values with the mode

data['Store'].fillna(data['Store'].mode()[0], inplace=True)
data['Product'].fillna(data['Product'].mode()[0], inplace=True)

Converting Data Types

# Convert Date to datetime type

data['Date'] = pd.to_datetime(data['Date'])

# Check the data types

print(data.dtypes)

Removing Outliers

import numpy as np

# Remove outliers based on Z-score

from scipy.stats import zscore

data = data[(np.abs(zscore(data[['Sales', 'Quantity',

'Price']])) < 3).all(axis=1)]
Feature Engineering

# Extract month and year from the Date

data['Month'] = data['Date'].dt.month
data['Year'] = data['Date'].dt.year

# Drop the original Date column

data.drop(columns=['Date'], inplace=True)

3. Exploratory Data Analysis (EDA)

import matplotlib.pyplot as plt

import seaborn as sns

# Sales over time

plt.figure(figsize=(10, 6))
sns.lineplot(data=data, x='Month', y='Sales', hue='Year')
plt.title('Monthly Sales Over Years')
plt.show()

# Sales by store
plt.figure(figsize=(10, 6))
sns.barplot(data=data, x='Store', y='Sales')
plt.title('Sales by Store')
plt.show()

# Sales by product
plt.figure(figsize=(10, 6))
sns.barplot(data=data, x='Product', y='Sales')
plt.title('Sales by Product')
plt.show()

4. Machine Learning Modeling

Splitting the Data

from sklearn.model_selection import train_test_split

# Define features and target variable

X = data.drop(columns=['Sales'])
y = data['Sales']

# One-hot encoding for categorical variables

X = pd.get_dummies(X, drop_first=True)

# Split the data into training and test sets

X_train, X_test, y_train, y_test = train_test_split(X, y,
test_size=0.2, random_state=42)

Training a Simple Model

from sklearn.linear_model import LinearRegression

from sklearn.metrics import mean_squared_error, r2_score

# Initialize the model

model = LinearRegression()

# Train the model

model.fit(X_train, y_train)

# Make predictions
y_pred = model.predict(X_test)

# Evaluate the model

mse = mean_squared_error(y_test, y_pred)
r2 = r2_score(y_test, y_pred)

print(f'Mean Squared Error: {mse}')

print(f'R^2 Score: {r2}')

5. Model Interpretation and Conclusion

You can interpret the model by looking at the coefficients:

coefficients = pd.DataFrame(model.coef_, X.columns,

columns=['Coefficient'])
print(coefficients)

Summary

In this guide, we performed the following steps:

 1. Loaded and cleaned the coffee sales data.
2. Conducted exploratory data analysis (EDA) to visualize sales trends.
3. Prepared the data for machine learning by handling categorical variables and
splitting the dataset.
4. Trained a simple linear regression model to predict sales.
5. Evaluated the model's performance.

This is a basic example. For a more robust analysis, you might consider advanced
techniques like cross-validation, feature selection, and trying different algorithms.

Sample code

Objective¶

This dataset contains detailed records of coffee sales from a vending machine. The dataset spans from
March 2024 to Present time, capturing daily transaction data. In this notebook, we are going to use EDA to
discover the customer's purchasing patterns and sales trends which can aid in the inventory planning.

Import packages

In [1]:

import numpy as np # linear algebra

import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import matplotlib.pyplot as plt
import seaborn as sns
import datetime as dt
import warnings
warnings.filterwarnings('ignore')

import os
for dirname, _, filenames in os.walk('/kaggle/input'):
for filename in filenames:
print(os.path.join(dirname, filename))

/kaggle/input/coffee-sales/index.csv

Load data

In [2]:

coffee_data = pd.read_csv('/kaggle/input/coffee-sales/index.csv')

EDA

In [3]:

coffee_data.head()

Out[3]:

cash_typ mone
date datetime card coffee_name
e y

2024-03-01
0 2024-03-01 card ANON-0000-0000-0001 38.7 Latte
10:15:50.520

2024-03-01 Hot
1 2024-03-01 card ANON-0000-0000-0002 38.7
12:19:22.539 Chocolate
 2024-03-01 Hot
2 2024-03-01 card ANON-0000-0000-0002 38.7
12:20:18.089 Chocolate

2024-03-01
3 2024-03-01 card ANON-0000-0000-0003 28.9 Americano
13:46:33.006

2024-03-01
4 2024-03-01 card ANON-0000-0000-0004 38.7 Latte
13:48:14.626

In [4]:

coffee_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1133 entries, 0 to 1132
Data columns (total 6 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 date 1133 non-null object
1 datetime 1133 non-null object
2 cash_type 1133 non-null object
3 card 1044 non-null object
4 money 1133 non-null float64
5 coffee_name 1133 non-null object
dtypes: float64(1), object(5)
memory usage: 53.2+ KB

In [5]:

coffee_data.isnull().sum()

Out[5]:

date 0
datetime 0
cash_type 0
card 89
money 0
coffee_name 0

dtype: int64

In [6]:

coffee_data.duplicated().sum()

Out[6]:

In [7]:

coffee_data.describe().T

Out[7]:

ma
count mean std min 25% 50% 75%
x

mone
1133.0 33.105808 5.035366 18.12 28.9 32.82 37.72 40.0
y

In [8]:

coffee_data.loc[:,['cash_type','card','coffee_name']].describe().T

Out[8]:

coun uniqu
top freq
t e
 cash_type 1133 2 card 1044

card 1044 446 ANON-0000-0000-0012 88

coffee_nam
1133 8 Americano with Milk 268
e

● There are 1033 transactions in the data.

● 89 missing values in the column 'card'.
● No duplicates.
● 2 unique values of 'cash_type'.
● 8 different coffee types with 'Americano with Milk' is the most popular product.

Let's check the transactions with missing value in 'card'.

In [9]:

coffee_data[coffee_data['card'].isnull()]['cash_type'].value_counts()

Out[9]:

cash_type
cash 89

Name: count, dtype: int64

All of the transactions with null 'card' information are from cash users.

In [10]:

coffee_data['cash_type'].hist()

Out[10]:

<Axes: >
 In [11]:

coffee_data['cash_type'].value_counts(normalize=True)

Out[11]:

cash_type
card 0.921447
cash 0.078553

Name: proportion, dtype: float64

~92% of the transactions are from card users.

In [12]:

pd.DataFrame(coffee_data['coffee_name'].value_counts(normalize=True).sort_values(asc
ending=False).round(4)*100)

Out[12]:
 proportio
n

coffee_name

Americano with
23.65
Milk

Latte 21.45

Cappuccino 17.30

Americano 14.92

Cortado 8.74

Hot Chocolate 6.53

Espresso 4.32

Cocoa 3.09

Americano with Milk and Latte are our most popular coffee products. In the second tier are Cappuccino
and Americano, while Cortado, Hot Chocolate, Espresso, and Cocoa are less popular.

Next, let's conduct data transformations for further analysis.

In [13]:

#Convert date and datetime to datetme format

coffee_data['date']=pd.to_datetime(coffee_data['date'])
coffee_data['datetime']=pd.to_datetime(coffee_data['datetime'])
#Create column of Month, Weekdays, and Hours
coffee_data['month']=coffee_data['date'].dt.strftime('%Y-%m')
coffee_data['day']=coffee_data['date'].dt.strftime('%w')
coffee_data['hour']=coffee_data['datetime'].dt.strftime('%H')

In [14]:

coffee_data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1133 entries, 0 to 1132
Data columns (total 9 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 date 1133 non-null datetime64[ns]
1 datetime 1133 non-null datetime64[ns]
2 cash_type 1133 non-null object
3 card 1044 non-null object
4 money 1133 non-null float64
5 coffee_name 1133 non-null object
6 month 1133 non-null object
7 day 1133 non-null object
8 hour 1133 non-null object
dtypes: datetime64[ns](2), float64(1), object(6)
memory usage: 79.8+ KB

In [15]:

coffee_data.head()

Out[15]:
 cash_ty mone coffee_nam da hou
date datetime card month
pe y e y r

2024-03-0 2024-03-01 ANON-0000-0000-00 2024-0

0 card 38.7 Latte 5 10
1 10:15:50.520 01 3

2024-03-0 2024-03-01 ANON-0000-0000-00 Hot 2024-0

1 card 38.7 5 12
1 12:19:22.539 02 Chocolate 3

2024-03-0 2024-03-01 ANON-0000-0000-00 Hot 2024-0

2 card 38.7 5 12
1 12:20:18.089 02 Chocolate 3

2024-03-0 2024-03-01 ANON-0000-0000-00 2024-0

3 card 28.9 Americano 5 13
1 13:46:33.006 03 3

2024-03-0 2024-03-01 ANON-0000-0000-00 2024-0

4 card 38.7 Latte 5 13
1 13:48:14.626 04 3

In [16]:

[coffee_data['date'].min(),coffee_data['date'].max()]

Out[16]:

[Timestamp('2024-03-01 00:00:00'), Timestamp('2024-07-31 00:00:00')]

The time range of this data set is from 2023-3-1 to 2024-7-31

Let's first check the overal revenue by products.

In [17]:
revenue_data =
coffee_data.groupby(['coffee_name']).sum(['money']).reset_index().sort_values(by='mo
ney',ascending=False)

In [18]:

plt.figure(figsize=(10,4))
ax = sns.barplot(data=revenue_data,x='money',y='coffee_name',color='steelblue')
ax.bar_label(ax.containers[0], fontsize=6)
plt.xlabel('Revenue')

Out[18]:

Text(0.5, 0, 'Revenue')

Latte is the product with the highest revenue, while Expresso is the one at the bottom. Then let's check
the monthly data.

In [19]:

monthly_sales =
coffee_data.groupby(['coffee_name','month']).count()['date'].reset_index().rename(co
lumns={'date':'count'}).pivot(index='month',columns='coffee_name',values='count').re
set_index()
monthly_sales
 Out[19]:

coffee_nam American Americano with Cappuccin Coco Cortad Espress Hot Latt
month
e o Milk o a o o Chocolate e

2024-0
0 36 34 20 6 30 10 22 48
3

2024-0
1 35 42 43 6 19 7 13 31
4

2024-0
2 48 58 55 9 17 8 14 58
5

2024-0
3 14 69 46 5 19 10 14 50
6

2024-0
4 36 65 32 9 14 14 11 56
7

In [20]:

monthly_sales.describe().T.loc[:,['min','max']]

Out[20]:

ma
min
x
coffee_name

Americano 14.0 48.0

Americano with
34.0 69.0
Milk

Cappuccino 20.0 55.0

Cocoa 5.0 9.0

Cortado 14.0 30.0

Espresso 7.0 14.0

Hot Chocolate 11.0 22.0

Latte 31.0 58.0

In [21]:

plt.figure(figsize=(12,6))
sns.lineplot(data=monthly_sales)
plt.legend(loc='upper left')
plt.xticks(range(len(monthly_sales['month'])),monthly_sales['month'],size='small')
 Out[21]:

([<matplotlib.axis.XTick at 0x7d45ae8a0430>,
<matplotlib.axis.XTick at 0x7d45ae8a0400>,
<matplotlib.axis.XTick at 0x7d45ae8a2ef0>,
<matplotlib.axis.XTick at 0x7d45ae8d3ee0>,
<matplotlib.axis.XTick at 0x7d45ae9149d0>],
[Text(0, 0, '2024-03'),
Text(1, 0, '2024-04'),
Text(2, 0, '2024-05'),
Text(3, 0, '2024-06'),

Text(4, 0, '2024-07')])

As shown in the line chart above, Americano with Milk and Latte, and Cappuccino are top selling coffee
types, while Cocoa and Expresso have lowest sales. Additionally, Americano with Milk and Latte show an
upward trending.

In [22]:

weekday_sales =
coffee_data.groupby(['day']).count()['date'].reset_index().rename(columns={'date':'c
ount'})
weekday_sales
 Out[22]:

da coun
y t

0 0 151

1 1 151

2 2 185

3 3 165

4 4 164

5 5 163

6 6 154

In [23]:

plt.figure(figsize=(12,6))
sns.barplot(data=weekday_sales,x='day',y='count',color='steelblue')
plt.xticks(range(len(weekday_sales['day'])),['Sun','Mon','Tue','Wed','Thur','Fri','S
at'],size='small')
 Out[23]:

([<matplotlib.axis.XTick at 0x7d45aea5b070>,
<matplotlib.axis.XTick at 0x7d45aea5b040>,
<matplotlib.axis.XTick at 0x7d45aea5af50>,
<matplotlib.axis.XTick at 0x7d45aeaa1240>,
<matplotlib.axis.XTick at 0x7d45aeaa1cf0>,
<matplotlib.axis.XTick at 0x7d45cf8c5f00>,
<matplotlib.axis.XTick at 0x7d45aeaa29b0>],
[Text(0, 0, 'Sun'),
Text(1, 0, 'Mon'),
Text(2, 0, 'Tue'),
Text(3, 0, 'Wed'),
Text(4, 0, 'Thur'),
Text(5, 0, 'Fri'),

Text(6, 0, 'Sat')])

The bar chart reveals that Tuesday has the highest sales of the week, while sales on the other days are
relatively similar.

In [24]:

daily_sales =
coffee_data.groupby(['coffee_name','date']).count()['datetime'].reset_index().reset_
index().rename(columns={'datetime':'count'}).pivot(index='date',columns='coffee_name
',values='count').reset_index().fillna(0)
daily_sales

Out[24]:

coffee_nam American Americano with Cappuccin Coco Cortad Espress Hot Latt
date
e o Milk o a o o Chocolate e

2024-03-0
0 1.0 4.0 0.0 1.0 0.0 0.0 3.0 2.0
1

2024-03-0
1 3.0 3.0 0.0 0.0 0.0 0.0 0.0 1.0
2

2024-03-0
2 1.0 2.0 0.0 1.0 2.0 0.0 2.0 2.0
3

2024-03-0
3 0.0 1.0 0.0 0.0 0.0 1.0 0.0 2.0
4

2024-03-0
4 0.0 0.0 0.0 1.0 1.0 0.0 4.0 3.0
5

... ... ... ... ... ... ... ... ... ...

2024-07-2
145 0.0 5.0 4.0 0.0 0.0 2.0 0.0 2.0
7
 2024-07-2
146 0.0 1.0 0.0 0.0 0.0 1.0 0.0 1.0
8

2024-07-2
147 3.0 2.0 2.0 1.0 0.0 0.0 2.0 1.0
9

2024-07-3
148 2.0 12.0 2.0 0.0 3.0 2.0 0.0 3.0
0

2024-07-3
149 2.0 6.0 1.0 2.0 4.0 0.0 0.0 7.0
1

150 rows × 9 columns

In [25]:

daily_sales.iloc[:,1:].describe().T.loc[:,['min','max']]

Out[25]:

mi ma
n x

coffee_name

Americano 0.0 5.0

 Americano with
0.0 12.0
Milk

Cappuccino 0.0 9.0

Cocoa 0.0 2.0

Cortado 0.0 4.0

Espresso 0.0 4.0

Hot Chocolate 0.0 4.0

Latte 0.0 7.0

This table provides us the infomation of how many of each products can be sold in each day.

In [26]:

hourly_sales =
coffee_data.groupby(['hour']).count()['date'].reset_index().rename(columns={'date':'
count'})
hourly_sales

Out[26]:
 hou coun
r t

0 07 13

1 08 44

2 09 50

3 10 133

4 11 103

5 12 87

6 13 78

7 14 76

8 15 65
9 16 77

10 17 77

11 18 75

12 19 96

13 20 54

14 21 70

15 22 35

In [27]:

sns.barplot(data=hourly_sales,x='hour',y='count',color='steelblue')

Out[27]:

<Axes: xlabel='hour', ylabel='count'>

Overall, two peak hours within each day can be observed: 10:00am and 7:00pm. Then, let's check if any
difference for different products.

In [28]:

hourly_sales_by_coffee =
coffee_data.groupby(['hour','coffee_name']).count()['date'].reset_index().rename(col
umns={'date':'count'}).pivot(index='hour',columns='coffee_name',values='count').fill
na(0).reset_index()
hourly_sales_by_coffee

Out[28]:

coffee_nam hou American Americano with Cappuccin Coco Espress Hot

Cortado Latte
e r o Milk o a o Chocolate
0 07 5.0 4.0 1.0 0.0 1.0 0.0 0.0 2.0

1 08 10.0 7.0 8.0 1.0 6.0 0.0 0.0 12.0

2 09 8.0 16.0 6.0 1.0 5.0 3.0 0.0 11.0

3 10 20.0 31.0 10.0 4.0 8.0 2.0 7.0 51.0

4 11 21.0 25.0 16.0 1.0 13.0 6.0 8.0 13.0

5 12 14.0 26.0 15.0 3.0 7.0 6.0 3.0 13.0

6 13 18.0 18.0 10.0 2.0 12.0 3.0 4.0 11.0

7 14 15.0 18.0 13.0 4.0 6.0 5.0 2.0 13.0

8 15 14.0 15.0 8.0 0.0 3.0 4.0 6.0 15.0

9 16 10.0 18.0 12.0 3.0 12.0 5.0 4.0 13.0

10 17 9.0 11.0 18.0 4.0 6.0 4.0 7.0 18.0

11 18 9.0 16.0 12.0 2.0 5.0 5.0 10.0 16.0

12 19 5.0 18.0 34.0 2.0 5.0 1.0 9.0 22.0

13 20 1.0 12.0 13.0 6.0 5.0 3.0 6.0 8.0

14 21 5.0 25.0 13.0 1.0 3.0 1.0 3.0 19.0

15 22 5.0 8.0 7.0 1.0 2.0 1.0 5.0 6.0

In [29]:

fig, axs = plt.subplots(2, 4, figsize=(20, 10))

# Flatten the array of subplots for easy iteration

axs = axs.flatten()

# Loop through each column in the DataFrame, skipping the 'Index' column
for i, column in enumerate(hourly_sales_by_coffee.columns[1:]): # Skip the first
column ('Index')
axs[i].bar(hourly_sales_by_coffee['hour'], hourly_sales_by_coffee[column])
axs[i].set_title(f'{column}')
axs[i].set_xlabel('Hour')
#axs[i].set_ylabel('Sales')

plt.tight_layout()

# Show the plot

plt.show()
The plots above illustrate the shopping traffic for each product throughout the day. Notably, all products
experience a peak in traffic around 10:00 AM, with this trend being particularly pronounced for Latte.
Additionally, Cappuccino, Cocoa, and Hot Chocolate tend to be more popular during the evening hours,
specifically between 6:00pm and 8:00pm.

Conclusion

From the analysis above, we have uncovered valuable insights into customer shopping patterns on a
daily and weekly basis. We have identified the most popular coffee products and observed the shopping
trends over time. These findings are instrumental in optimizing inventory planning, designing the layout
of vending machines, and determining the ideal restock times for coffee products.

1 Reference link
2 Reference link for ML project