📝 Release Description
We’re calling this v0.7.0 to reflect the maturity of the library’s preprocessing module, which delivers 7 essential functionalities. This is the first public release of dataglass, a Python library designed to streamline and standardize data cleaning before Exploratory Data Analysis (EDA) or machine learning.
🔧 Key Features in v0.7.0
❓ Missing Value Handling
- Drop rows
- Impute by datatype (mean, median, mode)
- Impute by adjacent values (forward/backward fill)
- Interpolation (linear, time-based)
📑 Duplicate Detection & Removal
- Detect and remove using full and partial similarity matching.
- Exact duplicates
- Fuzzy matching
❗ Outlier Detection & Handling
- Detection methods: IQR, Z-Score, Isolation Forest, Local Outlier Factor (LOF)
- Handling: Drop, replace with median, cap with boundaries
- Visualizations: Before vs. after (boxplot, histogram)
🔢 Feature Encoding
- Label encoding
- One-hot encoding
- Hashing
🔁 Type Conversion
- Automatic datatype inference
- User-defined conversion options
📏 Feature Scaling
- Min-Max scaling
- Z-Score (standard) scaling
- Robust scaling
- L2 normalization
🧩 Pipeline Compatibility
- Chain preprocessing steps using a custom pipeline class
💡 Additional Highlights
- Non-destructive processing: input data remains unchanged.
- Fully tested across Python 3.10–3.13.