Upload README.md

README.md

@@ -4,6 +4,11 @@ task_categories:
   - tabular-classification
 size_categories:
   - 100K<n<1M
+configs:
+  - config_name: default
+    data_files:
+      - split: train
+        path: paysim_fraud_cleaned_sample.csv
 ---
 
 # PaySim Financial Fraud Detection — EDA & Dataset Analysis
@@ -65,11 +70,11 @@ This project investigates whether patterns in transaction characteristics—part
 
 Applied **Interquartile Range (IQR) method** on `amount` feature. **5,358 outliers detected** (5.4% of data). Decision: retained all outliers because fraud transactions often occur at extreme values.
 
-![Outlier Detection — Amount Distribution](plots/eda_outlier_boxplot.png)
+![Outlier Detection — Amount Distribution](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/eda_outlier_boxplot.png)
 
 Applied **log1p transformation** to `amount` → `amount_log` for cleaner visualization.
 
-![Log-Transformed Amount Distribution](plots/eda_log_distribution.png)
+![Log-Transformed Amount Distribution](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/eda_log_distribution.png)
 
 ### 3. Descriptive Statistics
 
@@ -79,19 +84,19 @@ Calculated summary statistics by fraud status. Key insight: fraud transactions a
 
 Generated correlation heatmap of all numerical features to identify relationships between balance features and fraud status.
 
-![Correlation Heatmap](plots/eda_correlation_heatmap.png)
+![Correlation Heatmap](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/eda_correlation_heatmap.png)
 
 ### 5. Transaction Type Distribution
 
 Visualized how transactions are distributed across all 5 types.
 
-![Transaction Type Distribution](plots/eda_transaction_type_pie.png)
+![Transaction Type Distribution](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/eda_transaction_type_pie.png)
 
 ### 6. Fraud by Transaction Type (Raw Count)
 
 Showed where fraud actually occurs across transaction types before normalizing for volume.
 
-![Fraud Count by Transaction Type](plots/eda_fraud_by_type_bar.png)
+![Fraud Count by Transaction Type](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/eda_fraud_by_type_bar.png)
 
 ---
 
@@ -125,7 +130,7 @@ Showed where fraud actually occurs across transaction types before normalizing f
 
 **Finding:** Yes, fraud transactions occupy the extreme upper tail of the amount distribution. The highest-value transactions show disproportionately high fraud rates, indicating that transaction magnitude is a critical risk indicator.
 
-![Q1 — Amount Distribution by Fraud Status](plots/q1_amount_boxplot.png)
+![Q1 — Amount Distribution by Fraud Status](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/q1_amount_boxplot.png)
 
 **Conclusion:** Amount is a strong univariate fraud signal. Fraudsters tend to target high-value transfers to maximize stolen funds per transaction.
 
@@ -138,7 +143,7 @@ Showed where fraud actually occurs across transaction types before normalizing f
 - Legitimate transactions show account depletion in only **~0.3% of cases**
 - This is a massive logical inconsistency: legitimate users rarely empty accounts, but fraudsters consistently do
 
-![Q2 — Account Drain Rate: Fraud vs Legitimate](plots/q2_account_drain.png)
+![Q2 — Account Drain Rate: Fraud vs Legitimate](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/q2_account_drain.png)
 
 **Conclusion:** Balance depletion is one of the strongest fraud indicators in the dataset. This "emptying the well" behavior is a logical fingerprint of fraudulent activity — attackers maximize extraction and disappear.
 
@@ -151,7 +156,7 @@ Showed where fraud actually occurs across transaction types before normalizing f
 - PAYMENT, DEBIT, and CASH_IN have a 0% fraud rate in this dataset
 - Fraud is concentrated in transaction types that move money away from the original account
 
-![Q3 — Fraud Rate (%) by Transaction Type](plots/q3_fraud_rate_by_type.png)
+![Q3 — Fraud Rate (%) by Transaction Type](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/q3_fraud_rate_by_type.png)
 
 **Conclusion:** Transaction type is a critical categorical predictor. A simple rule-based filter (flag TRANSFER & CASH_OUT only) would catch 100% of fraud with minimal false positives.
 
@@ -166,7 +171,7 @@ Showed where fraud actually occurs across transaction types before normalizing f
 - Very Large (>100K): ~1.08% fraud rate
 - **Clear monotonic increase:** Larger amount buckets have exponentially higher fraud risk
 
-![Q4 — Fraud Rate (%) by Transaction Amount Range](plots/q4_fraud_by_amount_range.png)
+![Q4 — Fraud Rate (%) by Transaction Amount Range](https://huggingface.co/datasets/MichaelGelshtein/fraud-detection-eda/resolve/main/plots/q4_fraud_by_amount_range.png)
 
 **Conclusion:** Amount binning creates a simple but effective fraud scoring feature. The largest transactions contain ~8x more fraud than the smallest, making amount a strong standalone predictor.