User Guide

This guide gives you a topic-by-topic introduction to onTime, showcasing concrete use cases through runnable notebooks.

Note

This guide is a work in progress. If you have any questions or suggestions, please feel free to contact us.

The user guide is organized into three sections:

1. Core: the fundamental building blocks of the library.

2. Module: higher-level features built on top of the core, such as benchmarking and ML preprocessing.

3. Context: applied, real-world scenarios showing onTime in action.

Core

The core notebooks introduce onTime’s foundational objects — time series, detectors, generators, models, plots and processors — and how they fit together. Start here if you’re new to the library.

• Time Series
  • Get some data
  • Plot
• Time Series - Data Loading Helpers
  • From Darts to onTime
  • From Pandas to onTime
  • From .csv files to onTime
  • From regular data (e.g. Numpy) to onTime
• Detectors
  • Generate a TimeSeries
  • Detection given an abolute threshold
  • Detection given a statistical threshold (quantile)
  • Activate a logger when anomalies are detected
• Generators
  • Constant time series
  • Gaussian Noise TimeSeries
  • Random Walk Time Series
• Models
  • Models
  • Autoencoder
• Plots
  • Load data
  • Primitive Plots
  • Combined Plots
  • Subplots
  • Thematic Plots
• Processing
  • Generation of random time series
  • Apply function on the whole time series
  • Apply Function on Windows of the Time Series
  • Split Time Series in defined durations of e.g. day, week, month, year
  • Compute Correlation Through Time
  • Compute the Density of a TimeSeries
• Time Series
  • Generation of random time series
  • Use TimeSeries object
  • Custom Class Creation
  • Data Loading Helpers
• Fourier Processor
  • Generation of a signal with two frequencies
  • Apply the FFT on sliding windows
  • Visualize the frequency content over time
  • Restrict the frequency range

Module

The module notebooks cover features built on top of the core: data handling and datasets, anomaly frequency analysis, preprocessing for PyTorch and TensorFlow, and model benchmarking.

• Module - Data
  • Get dataset reference
  • Load the selected dataset
• Anomaly Frequency
• Processing
  • Module - Preprocessing
  • Module - Cross Validation
  • PyTorch Dataset
  • TensorFlow Dataset
• Module - Benchmarking
  • Initialization
  • Creating and running a Benchmark
  • Visualizing results

Context

The context notebooks show onTime applied to real-world, domain-specific scenarios rather than isolated features.

• Context - Common