Time Series & Forecasting
Predict what happens next: trend, seasonality, ARIMA, boosted trees, deep models, and today's pretrained forecasting models.
Before this: Machine Learning. Deep Learning helps for the later steps.
-
10 min
Step 1Stationarity
Why most classical models need a steady mean and variance.
Time Series Talk : Stationarityritvikmath
-
5 min
Step 2Trend, seasonality, and decomposition
Split a series into parts you can reason about.
What is Time Series DecompositionAric LaBarr
-
10 min
Step 3Exponential smoothing and Holt-Winters
A strong baseline that handles trend and seasons.
What is Holt Winters Method? - Time Series ForecastingEgor Howell
Nixtla/statsforecast★ 5k -
16 min
Step 4ARIMA
Autoregression, differencing, and moving averages in one model.
ARIMA Model Explained | Time Series ForecastingEgor Howell
-
11 min
Step 5Cross-validation for time series
Validate forward in time so you never peek at the future.
Cross-Validation for Time Series Forecasting | Python TutorialEgor Howell
-
23 min
Step 6Forecasting with XGBoost
Lag and calendar features with gradient-boosted trees.
Time Series Forecasting with XGBoost - Use python and machine learning to predict energy consumptionRob Mulla
-
30 min
Step 7LSTM forecasting
A first neural forecaster, with its limits shown clearly.
LSTM Time Series Forecasting Tutorial in PythonGreg Hogg
-
21 min
Step 8Temporal Fusion Transformer
Multi-horizon forecasts with attention and interpretable features.
Temporal Fusion Transformers, EXPLAINED. Advanced Time Series ForecastingData Heroes
Nixtla/neuralforecast★ 4.3k -
19 min
Step 9Time series foundation models
Why pretrained forecasters can work zero-shot.
Time Series Foundation ModelsAnastasia Borovykh
-
49 min
Step 10Chronos: pretrained forecasting
Forecast with a pretrained model and read its probabilistic output.
Chronos: Time series forecasting in the age of pretrained modelsAutoML Space
amazon-science/chronos-forecasting★ 6.3k -
34 min
Step 11Anomaly detection basics
Spot unusual points with robust statistics and isolation forests.
Anomaly detection in time series with Python | Data Science with MarcoData Science with Marco
Build it
Watching isn't learning. Pick one and put it on GitHub.
- Forecast daily demand for one public dataset with seasonal-naive, ARIMA, and XGBoost baselines, scored with rolling-origin validation.
- Compare a pretrained model such as Chronos with your best classical model on three series.
- Build an anomaly detector for sensor or server data using z-scores and Isolation Forest, and plot its alerts.