trend_classifier

Automated signal segmentation, trend classification and analysis.

Documentation | Tutorials | API Reference

Quick Start

from trend_classifier import Segmenter
seg = Segmenter(x=xdata, y=ydata, n=20) seg.calculate_segments() seg.plot_segments()

Installation

pip install trend-classifier

With optional dependencies:

pip install trend-classifier[pelt]         # PELT algorithm (ruptures)
pip install trend-classifier[optimization] # Hyperparameter tuning (optuna)
pip install trend-classifier[all]          # All extras

Features

• Multiple detection algorithms:

• Rich segment information: slope, offset, volatility, trend consistency
• DataFrame export: seg.segments.to_dataframe()
• Visualization: plotsegments(), plotsegment()
• Configurable: Fine-tune sensitivity with alpha, beta, window size

Example with Stock Data

import yfinance as yf
from trend_classifier import Segmenter
Download data
df = yf.download("AAPL", start="2020-01-01", end="2023-01-01", progress=False)
Segment and visualize
seg = Segmenter(df=df, column="Close", n=20)
seg.calculate_segments()
seg.plot_segments()
Export to DataFrame
seg.segments.to_dataframe()

Using Different Detectors

from trend_classifier import Segmenter
PELT algorithm (requires: pip install trend-classifier[pelt])
seg = Segmenter(x=x, y=y, detector="pelt", detector_params={"penalty": 10})
seg.calculate_segments()
Bottom-up with target segment count
seg = Segmenter(x=x, y=y, detector="bottomup", detectorparams={"max_segments": 10})
seg.calculate_segments()

Segment Properties

Each segment contains:

| Property | Description | |----------|-------------| | start, stop | Index range | | slope | Trend direction and steepness | | std | Volatility (after detrending) | | reasonfornew_segment | Why segment boundary was placed |

segment = seg.segments[0]
print(f"Slope: {segment.slope:.4f}, Volatility: {segment.std:.4f}")

Documentation

Full documentation with tutorials and API reference:

https://izikeros.github.io/trend_classifier/

License

MIT © Krystian Safjan