aa-prepflow 0.1.1

No project description provided

AA Library Preparation Flow

Integrated preprocessing tool combining an interactive UI/UX interface and a modular Python library to simplify data cleaning and exploration — allowing both visual experimentation and direct implementation to ensure efficient, consistent, and scalable preprocessing workflows.

1. Installation & Import

pip install aa_prepflow

import pandas as pd
import numpy as np
from aa_prepflow import AAPrepflowCrossSection, AAPrepflowTimeSeries, AAPrepflowPanel, AAPrepflowBase

---

2. API Function

A. Direct Method (Quick)

Use this method for quick exploration. This method combines fit and transform in a single call.

# 1. Initialize Flow based on data type
flow_cs = AAPrepflowCrossSection(data_cs)
flow_ts = AAPrepflowTimeSeries(data_ts, col_time='date')
flow_panel = AAPrepflowPanel(data_panel, col_time='date', group_by='city')

# 2. Handling Missing Value (replace <method_name> from the table below)
data_cleaned = flow_cs.mv.clean_<method_name>()

# 3. Handling Outlier (replace <method_name> from the table below)
data_cleaned = flow_ts.outlier.apply_<method_name>()

B. Fit/Transform Method (Production Ready)

Use this pattern for production pipelines so you can fit on training data and transform on both training and testing data separately.

# 1. Initialize Flow with TRAINING data
flow = AAPrepflowTimeSeries(data_train_ts, col_time='date')

# 2. FIT on TRAINING data
# Learn parameters from train data (e.g.: median, IQR bounds, etc.)
flow.mv.fit_<method_name>()
flow.outlier.fit_<method_tame>()

# 3. TRANSFORM on Train and Test data
# Apply the learned parameters

# --- Transform Train Data ---
# (Using internal fitted data)
cleaned_train = flow.mv.transform()
cleaned_train = flow.outlier.transform_<handler_name>(cleaned_train)

# --- Transform Test Data ---
# (Using external test data)
cleaned_test = flow.mv.transform(test_data)
cleaned_test = flow.outlier.transform_<handler_name>(cleaned_test)

---

3. API Interactive Lab (Gradio)

You can launch an interactive lab to visually test strategies.

A. Lab with DataFrame

This method will directly load your data into the Gradio application.

# Initialize with data, group_by, and col_time
lab = AAPrepflowBase(data=df_panel, group_by='City', col_time='Date')

# Launch Interactive Lab
lab.flow_lab()

B. Lab without DataFrame

This method will launch an empty Gradio application, and you can upload a CSV file from the browser.

# Initialize AAPrepflowBase
lab = AAPrepflowBase()

# Launch Interactive Lab
lab.flow_lab()

---

4. API Reference <method_name>

Use the method names from the following table to replace <...> in the code examples above.

A. Missing Value (.mv)

Data Type	Method (Quick)	Method (Fit)	Strategy Description
Cross-Section	clean_listwise_deletion	fit_listwise_deletion	Remove rows with NA.
	clean_mean_median_mode_imputer	fit_mean_median_mode_imputer	Fill with mean, median, or mode.
	clean_hot_deck_imputer	fit_hot_deck_imputer	Fill with a random sample.
	clean_regression_imputer	fit_regression_imputer	Fill with regression prediction.
	clean_knn_imputer	fit_knn_imputer	Fill with K-Nearest Neighbors.
Time-Series / Panel	clean_locf	fit_locf	Last Observation Carried Forward.
	clean_nocb	fit_nocb	Next Observation Carried Backward.
	clean_interpolation	fit_interpolation	Interpolation (e.g.: linear).
	clean_seasonal_imputer	fit_seasonal_imputer	Fill with seasonal average.
	clean_arima_imputer	fit_arima_imputer	Fill with ARIMA prediction.
	clean_kalman_imputer	fit_kalman_imputer	Fill with Kalman Filter.
	clean_moving_average_imputer	fit_moving_average_imputer	Fill with MA average.
	clean_cagr_imputer	fit_cagr_imputer	Fill with geometric interpolation.

B. Outlier (.outlier)

Step 1: Detection (Quick apply_... or Fit fit_...)

Data Type	Method (Quick)	Method (Fit)	Detection Description
Cross-Section	apply_iqr_...	fit_iqr	Interquartile Range (IQR).
	apply_zscore_...	fit_zscore	Z-Score (Standard Deviation).
Time-Series / Panel	apply_iqr_...	fit_iqr	Interquartile Range (IQR).
	apply_zscore_...	fit_zscore	Z-Score (Standard Deviation).
	apply_rolling_iqr_...	fit_rolling_iqr	Rolling IQR (adaptive).

Step 2: Handling (Quick ..._capping or Transform transform_...)

Method (Quick)	Method (Transform)	Handling Description
apply_..._capping	transform_capping	Replace outlier with the boundary value.
apply_..._imputation	transform_imputation	Replace outlier with 'mean'/'median'.
apply_..._locf	transform_locf	Replace outlier with the last valid value.