Create an Outlier Detection and Handling Framework for Your Data

You are a statistical analysis expert. I have a [DATASET_TYPE] dataset with [NUMBER_OF_ROWS] records containing the following numerical features: [LIST_OF_NUMERICAL_COLUMNS]. The data will be used for [ANALYSIS_PURPOSE] in the [INDUSTRY] domain. Design a thorough outlier detection and handling framework: **Phase 1 — Detection** Apply and compare at least 4 outlier detection methods to my data: - Statistical methods (Z-score, Modified Z-score, IQR) - Visualization-based (box plots, scatter plots, distribution plots) - Machine learning-based (Isolation Forest, DBSCAN, Local Outlier Factor) For each method, provide: - Python implementation code - How to interpret the results - Strengths and limitations for [DATASET_TYPE] data **Phase 2 — Classification** Help me distinguish between: - True anomalies (errors or fraud) - Natural extreme values (legitimate but rare) - Domain-specific acceptable outliers in [INDUSTRY] Provide a decision tree or flowchart logic for classifying detected outliers. **Phase 3 — Handling** For each outlier category, recommend an appropriate action: - Removal, capping/winsorization, transformation (log, Box-Cox), imputation, or separate modeling - Justify each recommendation and explain its impact on [ANALYSIS_PURPOSE] **Phase 4 — Documentation** Create a summary report template that logs: total outliers found per method, classification decisions, actions taken, and before/after distribution statistics. Provide all code in Python using pandas, scipy, and scikit-learn.