nikolai-neustroev/tabular-playground-series-jan-2021

tabular-playground-series-jan-2021's Issues

Experiment 1 - PCA for Dimensionality Reduction

Principal Component Analysis for Dimensionality Reduction. Reduce the dimensionality of the data to n principal components.

Experiment 13 - Deep feature synthesis and feature selection

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-13

Experiment 7 - large min_child_samples

Experiment 14 - Add selected features

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-14

Experiment 4 - Two training sets

Experiment 3 - SVM Regression

Experiment 2 - Scale but no dim reduction

Experiment 8 - Drop outliers

Experiment 5 - Clean data and GMM Regression

https://github.com/AlexanderFabisch/gmr

Experiment 11 - Feature Generation by Division

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-11

Experiment 10 - Scaling

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-10

Experiment 9 - Log and Exp Transform

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-9

Experiment 12 - Feature selection

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-12

Experiment 15 - DFS and PCA

https://github.com/nikolai-neustroev/tabular-playground-series-jan-2021/tree/exp-15

Experiment 6 - Decision Tree

DecisionTreeClassifier (только нужно указать параметр min_samples_leaf достаточно большой, например len(X) / n_bins).
Деревья делают почти тоже самое, что и описанный выше алгоритм на гистограммах: строят границы бинов (пороги сплитов дерева - это и есть границы бинов), только делают это итерактивно, добавляя границы по одной.