Heartrate Anomaly Detection using Matrix Profiling

This project implements anomaly detection on heartrate time-series data using the Matrix Profile technique, powered by the STUMPY library. It reads input data, computes matrix profiles, identifies anomalies, visualizes results, and exports findings for further analysis.

Features

• Matrix Profile Computation: Detects unusual patterns in time-series data using STUMPY.
• Anomaly Detection: Identifies timestamps where anomalies occur based on a configurable threshold.
• Data Normalization: Optionally normalizes data to improve detection consistency.
• Interactive Plotting: Visualizes time series and anomalies with Plotly.
• Results Export: Saves computed profiles, scores, and anomalies to a pickle file.

Requirements

• Python 3.8+
• Required packages:
  • numpy
  • pandas
  • stumpy
  • plotly
  • pickle (standard library)
  • functools (standard library)
  • datetime (standard library)

Install dependencies using:

pip install numpy pandas stumpy plotly

Usage

1. Place your heartrate data as a CSV file in the data directory (default expects heartrate_personal_reduced.csv).
2. Run the script:

python src/matrix-profiling.py

3. The script will:
   • Read the input data
   • Compute matrix profiles
   • Detect anomalies using the specified threshold
   • Display plots for each column
   • Export results to mp_exported.pkl

Configuration

You can adjust parameters directly in the script:

• window_size: Sliding window size for matrix profile computation
• smooth_n: Rolling window size for smoothing
• normalize_data: Boolean to enable/disable normalization
• plot_results: Boolean to enable/disable plotting
• threshold: Float to set the anomaly detection threshold

Output

• Printed Output: Displays top scores and detected anomalies.
• Exported Pickle File: Contains:
  • mp_dists: Matrix profile distances per column
  • df_scores: Scores and rankings
  • anomalies: DataFrame with detected anomalies

Example

The script outputs something like:

            score  rank
heartrate    0.12   1.0

   timestamp     column    value  index  mp_score
0 2023-05-01   heartrate   85.0     150     0.95
1 2023-05-02   heartrate   120.0    320     1.10

License

This project is provided as-is for research and educational purposes.