Introduction to insss

insss is a Python package for detecting anomalous time series data using a variety of methods. It provides various algorithms for detecting anomalies in time series data, including statistical methods, machine learning approaches, and hybrid methods.

Features

• Multiple anomaly detection algorithms including ARIMA, Isolation Forests, and SVM.
• Support for both univariate and multivariate time series data.
• Ability to handle large scale time series data using various data preprocessing techniques.
• Easy to use with a simple API for model building and prediction.
• Provides visualization tools to easily interpret the results.

Installation

To install insss, you can use pip:

pip install insss

Usage

To use insss, you start by importing the desired detector:

from insss.detectors import ARIMADetector

Next, you create an instance of the chosen detector and fit it to your data:

detector = ARIMADetector()
detector.fit(train_data)

Finally, you can use the detector to make predictions on new data:

anomaly_scores = detector.predict(test_data)

Examples

Here's an example of how to use insss to detect anomalies in a univariate time series:

from insss.detectors import ARIMADetector
from insss.utils import generate_synthetic_data, plot_anomalies

# Generate synthetic data for testing
data = generate_synthetic_data(period=50, n=1000, anomaly_ratio=0.05)

# Split data into train and test sets
train_data = data[:800]
test_data = data[800:]

# Fit an ARIMA detector to the training data
detector = ARIMADetector()
detector.fit(train_data)

# Predict anomalies in the test data
anomaly_scores = detector.predict(test_data)

# Plot the data with detected anomalies
plot_anomalies(test_data, anomaly_scores)

This will produce a plot showing the original time series data with the detected anomalies highlighted.

Conclusion

insss provides a comprehensive set of tools for detecting anomalies in time series data. With its support for a variety of algorithms, easy-to-use API, and visualization tools, insss is a great choice for anyone looking to find anomalous patterns in their data.