Prediction algorithms
Plain-language guides to the algorithms behind Clarex predictions — what each one does, how it works, and when it's the right call.
Forecasting
Project a single time series forward over a horizon.
Choosing a forecasting algorithmLightGBM forecasting
Gradient-boosted quantile regression for time-series forecasting — a point prediction plus a calibrated uncertainty band for every period in the horizon.
AutoARIMA
Classical seasonal ARIMA (SARIMAX) for a single time series — automatic model search, a point forecast, and statistically-derived uncertainty bands.
Regression
Predict a numeric value from feature columns.
Choosing a regression algorithmLightGBM regression
Gradient-boosted regression for predicting a numeric value from feature columns — strong accuracy with little tuning and native categorical support.
Ridge regression
An interpretable linear model for predicting a numeric value — least squares with L2 regularization and readable, comparable coefficients.
Linear regression (OLS)
Textbook ordinary least squares for predicting a numeric value — the simplest interpretable baseline, with coefficients that are each feature's raw effect.
Random forest regression
A non-linear ensemble of decision trees for predicting a numeric value — robust, captures feature interactions, and works well with little tuning.
Classification
Predict a categorical label from feature columns.
Choosing a classification algorithmLightGBM classification
Gradient-boosted classification for predicting a categorical label from feature columns — strong out-of-the-box accuracy, native categorical handling, binary and multi-class.
Logistic regression
An interpretable linear classifier for predicting a categorical label from feature columns — fast, transparent, with readable coefficients.
Random forest classifier
A non-linear ensemble of decision trees for predicting a categorical label — robust, captures feature interactions, and works well with little tuning.