Прогнозирование временных рядов

Уровень 0 · материалов: 6

В кластер входят документы, посвященные техническим методам, библиотекам и теоретической классификации моделей прогнозирования временных рядов.

Общие признаки: методы прогнозирования временных рядов, сравнение статистических и ML моделей, практическая реализация на Python, стационарность данных

Группа выше: Прикладные задачи машинного обучения

Материалы

• #23369 — Time Series Analysis and Forecasting using Python — Представитель этого кластера

  Открытый курс машинного обучения. Тема 9. Анализ временных рядов с помощью Python — DataWondering

  Смысл: The main idea is to provide a practical roadmap for time series forecasting in Python, contrasting classical statistical methods (like Holt-Winters and SARIMA) with modern machine learning approaches (like XGBoost) while stressing the importance of data stationarity and proper validation.

  A practical guide to time series forecasting in Python, covering everything from basic smoothing and SARIMA to XGBoost and feature engineering.

• #18165 — Time Series Analysis Using Python and Statsmodels

  Анализ временных рядов с помощью python — kuznetsovin

  Смысл: The text serves as a practical demonstration of the end-to-end pipeline for time series forecasting in Python, specifically showing how to transform raw, non-stationary data into a stationary series to build and validate an ARIMA model.

  A step-by-step tutorial on analyzing non-stationary warehouse turnover data and building an ARIMA forecast model using Python's statsmodels and pandas libraries.

• #24408 — Predicting the Future with Facebook Prophet Library

  Предсказываем будущее с помощью библиотеки Facebook Prophet — miptgirl

  Смысл: The main idea is to introduce the Facebook Prophet library for time series forecasting, demonstrating its ease of use and functionality through a practical example, while honestly comparing its performance against classical statistical models like SARIMA.

  An introductory guide to using the Facebook Prophet library for time series forecasting, featuring a practical application on Habr post data and a performance comparison with SARIMA.

• #16133 — Overview of Time Series Forecasting Models

  Обзор моделей прогнозирования временных рядов: проба пера — mbureau

  Смысл: The text aims to categorize and evaluate the effectiveness of various time series forecasting models, distinguishing between widely applicable statistical tools and those the author considers academically pretentious but practically ineffective.

  A critical survey of time series forecasting models that categorizes mainstream methods and questions the practical utility of more obscure algorithmic approaches.

• #10390 — Time Series Forecasting with Recurrent Neural Networks

  Прогнозирование временных рядов с помощью рекуррентных нейронных сетей — OLZ1

  Смысл: The main idea is to provide a practical, step-by-step tutorial on how to implement time-series forecasting using LSTM networks in TensorFlow, progressing from simple one-variable models to complex multi-variable models that predict future sequences.

  A comprehensive technical guide on using TensorFlow's LSTM networks to forecast weather data through univariate and multivariate time-series analysis.

• #33383 — Classification of Forecasting Methods and Models

  Классификация методов и моделей прогнозирования — mbureau

  Смысл: The text aims to provide a systematic taxonomy for forecasting by separating the process (methods) from the mathematical tool (models) and further categorizing them by their nature (intuitive vs. formal) and their application (domain-specific vs. universal time series).

  A comprehensive breakdown of forecasting into intuitive and formalized methods, and domain-specific versus statistical or structural time series models.