PYTHON DATA SCIENCE PROJECT REPORT

BY SHIVAM
 ABSTRACT
This code project focuses on a comprehensive exploratory data analysis (EDA)
of a banking dataset. Utilizing Python's powerful data analysis and
visualization libraries, the project aims to extract meaningful insights and
present them through visual representations. The primary libraries used
include Pandas, NumPy, Matplotlib, and Seaborn. The analysis covers various
aspects of the dataset, including descriptive statistics, value counts, box plots,
and a correlation matrix heatmap.

Outcome:

• The project delivers a detailed exploratory data analysis that identifies

key characteristics and relationships within the banking dataset.
• Visualizations such as box plots and heatmaps help in identifying
trends, anomalies, and correlations.
• Specific insights include the distribution of ages, job types, marital
status, education levels, balance, and more.

Applications:

• Insights derived from this analysis can be instrumental in customer

segmentation, risk assessment, and marketing strategies.
• Banks can use these insights to tailor their services, improve customer
satisfaction, and enhance decision-making processes.
 INTRODUCTION
In the modern banking sector, data analysis plays a crucial role in understanding
customer behavior, managing risks, and devising strategic business decisions.
This project aims to perform an exploratory data analysis (EDA) on a banking
dataset to extract meaningful insights and visualize key aspects of the data. By
leveraging Python's robust data analysis libraries such as Pandas, NumPy,
Matplotlib, and Seaborn, this project provides a comprehensive overview of the
dataset, highlighting trends, relationships, and anomalies.

Objectives:

The primary objectives of this project are:

1. Data Loading and Preparation: Efficiently load and prepare the dataset for
analysis.

2.Descriptive Statistics: Generate summary statistics to understand the central

tendencies and distribution of the data.

3. Value Counts: Identify the frequency of unique values in categorical variables.

4. Visualization: Create visual representations, such as box plots and heatmaps,

to illustrate data distributions and correlations.

5. Detailed Analysis: Perform a thorough examination of key variables such as

age, job, marital status, education, default status, balance, loan status, contact
methods, and campaign outcomes.

6. Correlation Analysis: Explore the relationships between numerical variables

using a correlation matrix.
Methodology:

1. Data Loading:

- The dataset is imported using the Pandas library, which provides flexible and
powerful data structures for data manipulation.

2. Descriptive Statistics and Value Counts:

- Descriptive statistics such as mean, median, standard deviation, and quartiles

are calculated for numerical variables.

- Value counts are computed for categorical variables to understand their

distribution.

3. Visualization:

- Box plots are created for variables like age and job to visualize their
distribution and identify potential outliers.

- A correlation matrix heatmap is generated to explore the relationships between

numerical variables, providing insights into how they are interrelated.

4. Detailed Column Analysis:

- Each key variable is analyzed individually to extract specific insights. For

example, the analysis of the `age` column includes unique values, descriptive
statistics, value counts, and a box plot.

5. Correlation Matrix:

- A correlation matrix is created for numerical variables to identify significant

correlations, which are visualized using a heatmap for easy interpretation.
Dataset Description:

The dataset includes various attributes related to bank clients and their
interactions with the bank. Key variables include:

- Age: Age of the client.

- Job: Type of job the client has.

- Marital Status: Marital status of the client.

- Education: Educational background of the client.

- Default: Whether the client has credit in default.

- Balance: Average yearly balance in the client's account.

- Housing: Whether the client has a housing loan.

- Loan: Whether the client has a personal loan.

- Contact: Communication type used to contact the client.

- Day: Last contact day of the month.

- Month: Last contact month of the year.

- Duration: Duration of the last contact in seconds.

- Poutcome: Outcome of the previous marketing campaign.

- y: Whether the client subscribed to a term deposit.

Importance of Detailed Analysis in Banking:

Detailed Analysis is a crucial first step in any data analysis process, especially in
the banking sector, where understanding customer behaviour and financial
patterns is vital. This project not only aims to uncover hidden patterns and
anomalies in the dataset but also sets the stage for more advanced predictive
modelling and decision-making processes. By providing a clear and
comprehensive view of the data, Detailed Analysis helps banks to:

- Segment customers effectively.

- Assess and mitigate risks.

- Design targeted marketing strategies.

- Develop tailored financial products.

- Improve customer satisfaction and retention.

- Ensure compliance and detect fraud.

 PROJECT DESIGN
1. Import libraries:

• pandas (pd): for data manipulation

• numpy (np): for numerical computations
• matplotlib.pyplot (plt): for creating plots
• seaborn (sns): for creating statistical graphics

2. Load the data:

• Reads a CSV file named "banking_data.csv" located on your computer and

stores it in a pandas dataframe named "df".

3. Analyze individual columns:

• Loops through various columns in the dataframe and performs different

analysis on each:
o 'age': Gets unique values, descriptive statistics, value counts, and
creates a box plot.
o 'job': Gets value counts and creates a box plot.
o 'marital status': Gets value counts.
o 'education': Gets value counts.
o 'default': Gets value counts, descriptive statistics, calculates
proportion of defaults, and describes it.
o 'balance': Gets value counts, descriptive statistics.
o Similar analysis is done for other columns like 'housing', 'loan',
'contact', 'day', 'month', 'duration', 'poutcome', and 'y'.

4. Correlation Matrix:

• Selects only numerical columns from the data frame.

 • Calculates the correlation matrix which shows the correlation coefficients
between each pair of numerical columns.
• Creates a heatmap using seaborn to visualize the correlation matrix.
Heatmap uses colours to represent the strength and direction of the
correlations.

Functions used in the code:

1. pandas functions:

• pd.read_csv(filepath): This function reads data from a comma-

separated values (CSV) file located at the specified filepath and returns a
pandas dataframe object.
• df.select_dtypes(include=data_types): This function
selects columns from a pandas dataframe based on their data types. Here,
it selects only columns with data types 'int64' (integers) and 'float64'
(floating-point numbers).
• .unique(): This method applied to a pandas Series (representing a
single column) returns all unique values within that column.
• .describe(): This method applied to a pandas Series returns summary
statistics about the data in that column, like count, mean, standard
deviation, etc.
• .value_counts(): This method applied to a pandas Series returns the
number of times each unique value appears in that column.

2. matplotlib.pyplot functions:

• plt.figure(figsize=(width, height)): This function creates

a new figure window for plotting with the specified width and height.
• sns.boxplot(y='column_name', data=dataframe): This
function from seaborn, which is built on top of matplotlib, creates a box
 plot to visualize the distribution of data in a specified column (y) of a
pandas dataframe (data).
• plt.title("title_text"): This function sets the title for the
current plot.
• plt.show(): This function displays the currently created plots.

3. seaborn functions:

• sns.heatmap(data_matrix, annot=True,
cmap='color_scheme', fmt=".2f"): This function creates a
heatmap visualization for a correlation matrix (data_matrix). Here,
annot=True displays the correlation values within each cell,
cmap='PuBuGn' sets the colour scheme for the heatmap, and fmt=".2f"
formats the displayed values to have 2 decimal places.

4. other functions:

• .corr(): This method applied to a pandas dataframe calculates the

correlation coefficient between each pair of numerical columns and
returns a correlation matrix as a data frame.
 OUTPUT

Age:

• Lower Edge of the Box: The first quartile (Q1).

• Line Inside the Box: The median (Q2).
• Upper Edge of the Box: The third quartile (Q3).
Job:
Marital Status: We can see, the marital status of our clients here.

Education: We can see the Educational qualifications of our clients here.

Default Credit: We can see here whether the clints have credit in default or
not.

Balance: The average yearly balance in euros for the clients.

Housing Loan: Clients who have taken housing loan.

Personal Loan: Clients who have taken personal loan.

Contact: The type of communication used to contact the client.

Contact Day: The last contact month of the year.

Outcome: The outcome of the previous marketing campaign.

Subscription: Indicates whether the client has subscribed to a term deposit.

Co-relation Matrix:

Interpretation of the co-relation matrix:

Correlation Coefficients:

• 1: Perfect positive correlation (as one variable increases, the other

increases proportionally).
• -1: Perfect negative correlation (as one variable increases, the other
decreases proportionally).
• 0: No correlation (no linear relationship between the variables).

Heatmap Colors:
 • Darker shades indicate stronger correlations (close to 1 or -1).
• Lighter shades indicate weaker correlations (close to 0).

Age and Balance:

• If the correlation coefficient is 0.3, it suggests a weak positive relationship.

Duration and Balance:

• If the correlation coefficient is 0.6, it indicates a moderate positive relationship,

meaning higher duration calls are somewhat associated with higher balances.

Day and Month:

• If the correlation coefficient is close to 0, it indicates no significant linear relationship

between the day of the month and the month of contact.
 CONCLUSION

The exploratory data analysis (EDA) performed on the banking dataset provides
valuable insights into the demographic and financial characteristics of the clients,
as well as the effectiveness of past marketing campaigns.

Key Findings:

- Client Demographics: The dataset reveals diverse age groups and job
categories, with varying marital statuses and education levels.

- Financial Behaviour: The analysis of account balances and loan statuses

indicates the financial health and risk profiles of the clients.

- Marketing Effectiveness: The outcomes of previous marketing campaigns and

the distribution of contact methods highlight areas for improvement in future
campaigns.

- Risk Assessment: The proportion of clients with defaulted credit and the
correlation matrix help in identifying potential risk factors and interrelationships
between features.

Future Prospects:

1. Targeted Marketing:

- Utilize the demographic and financial insights to create personalized

marketing strategies aimed at specific client segments.

- Focus on the most effective contact methods and times to improve campaign
success rates.
2. Risk Management:

- Implement more robust risk assessment models using the identified key
features (e.g., age, balance, loan status) to minimize defaults.

- Use the correlation matrix to refine predictive models by addressing

multicollinearity issues.

3. Product Development:

- Develop new financial products tailored to the needs of different client

demographics, such as age-specific savings plans or job-specific loan products.

- Enhance existing products based on client feedback and financial behaviour

patterns.

4. Predictive Modelling:

- Build predictive models to forecast client behaviours, such as the likelihood

of subscribing to a term deposit or defaulting on a loan.

- Use machine learning techniques to identify patterns and trends that can
inform strategic decisions.

5. Customer Segmentation:

- Leverage clustering techniques to segment clients into distinct groups based

on their demographics, financial behaviour, and past interactions.

- Tailor services and communications to each segment to enhance customer

satisfaction and retention.