Financial Transactions Data Science Project Using Pandas | EPISODE #2
Financial Transactions Dataset: https://www.kaggle.com/datasets/compu...
_____________________________________
GitHUB: https://github.com/mayankcodezzz/Fina...
_____________________________________
combined analysis: https://github.com/mayankcodezzz/Fina...
_____________________________________
1. Introduction:
This analysis explores relationships between user demographics, financial data, and card-related attributes. It involves data preprocessing, merging datasets, and statistical tests such as Chi-Square tests, ANOVA, and regression analysis to uncover patterns.
2 Data Preprocessing:
The dataset consists of user and card-related information. Key preprocessing steps include:
Dropping unnecessary columns (e.g., 'card on dark web').
Computing `current age` using birth year and month.
Converting financial values (e.g., `credit limit`, `yearly income`) from strings to numeric.
Converting `acct open date` to datetime format.
Determining `retirement status` based on age and retirement threshold.
3. Merging Datasets:
The user dataset was merged with the card dataset on `id` and `client id` to create a comprehensive dataset for analysis. The resulting dataset had:
6146 unique records.
No duplicated rows.
Various categorical and numerical features, including income, credit score, and card types.
4. Chi-Square Tests for Categorical Relationships:
Chi-Square tests were conducted to determine whether categorical variables were related.
4.1 Gender and Card Brand:
Null Hypothesis (H₀): Gender and card brand are independent.
Alternative Hypothesis (H₁): Gender affects card brand choice.
Result: P-value = 0.104, failing to reject H₀. No significant association found.
4.2 Gender and Card Type
Null Hypothesis (H₀): Gender and card type are independent.
Alternative Hypothesis (H₁): Gender affects card type choice.
Result: P-value = 0.404, failing to reject H₀. No significant association found.
5. Analysis of Variance (ANOVA) for Income Differences:
To determine if `yearly income` varies across different card brands:
Shapiro-Wilk test: Showed non-normality for all groups.
Levene’s test:: Showed equal variance (p = 0.3985).
Kruskal-Wallis test: p = 0.0267, rejecting H₀, indicating a significant difference in median income across card brands.
Dunn’s Post-hoc test: Significant income differences between certain card brands.
6. Regression Analysis:
A simple linear regression was conducted to explore the relationship between `yearly income` and `credit limit`.
Null Hypothesis (H₀): No relationship exists between yearly income and credit limit.
Alternative Hypothesis (H₁): A relationship exists.
Findings: The regression results suggest a statistically significant positive relationship between income and credit limit, meaning higher incomes generally lead to higher credit limits.
7. Conclusions
Gender does not significantly impact card brand or type selection.
Significant differences exist in yearly income based on card brand.
Yearly income is a significant predictor of credit limit.
Further analysis could involve additional variables or a more complex regression model.
This report provides key insights into user demographics, financial behavior, and their influence on card-related attributes. Let me know if you need any modifications or additional insights!
_____________________________________
#datascience #machinelearning #bankproject #datasciencebankproject #machinelearningprojectwithbank #financeprojectdatascience #financeproject #histograms #boxplots #polars