Some extra evaluation metrics for xgboost in R

100%

Some extra evaluation metrics for xgboost in R

✓ Published0🌍 Public

Last edited May 26, 2016

Created on May 26, 2016

This example demonstrates two custom evaluation metrics for the xgboost package in R: MAPE (mean absolute percentage error) and RMSEE (root mean square exponential error). It defines two R functions, `MAPE` and `RMSEE`, each of which takes prediction values and a training data object, then extracts labels via `getinfo` to compute the respective error metric. The code shows how to integrate custom metrics into xgboost’s training loop by returning a list with the metric name and computed value, enabling users to monitor these alternatives during model fitting.

AI-generated description

Similar vizzes

Loading thumbnail…

Xgboost cross validation functions for time series data + gridsearch functions in R

This example shows a custom cross-validation function for XGBoost models on time-series data, using explicitly defined training and test index lists rather than random folds. The R code implements `xgb.ts.cv`, which extends the standard `xgb.cv` API by accepting separate `traininds` and `testinds` parameters, enabling chronological splits. It internally uses `xgb.ts.cv.mkfold` to create fold-specific boosters and watchlists, then iterates through boosting rounds with `xgb.iter.update` and `xgb.iter.eval`, aggregating results via `xgb.cv.aggcv`. The function supports early stopping, custom metrics, and prediction output, mirroring the original xgboost package’s cross-validation workflow.

Aaakansh9

58% match

Loading thumbnail…

Custom Xgboost functions for Safe Aging with SPHERE Challenge

This R script implements custom cross-validation and training functions for XGBoost, targeting a 20-column regression problem from the SPHERE challenge on safe aging. It trains 20 separate gradient-boosted tree models simultaneously, one per target column, using `xgb.DMatrix` and the internal `xgb.iter.update` API to update all boosters in lockstep. The cross-validation function computes a weighted Brier score across all targets, normalizes predictions per row, and supports early stopping. The code relies on the `xgboost` and `dplyr` packages, with prediction output as a data frame.

Aaakansh9

56% match

Loading thumbnail…

Common feature exploration functions for Data Science Competitions in R

This R script provides a collection of feature-exploration functions for data science competitions, demonstrating common preprocessing tasks. It shows how to remove duplicate and highly correlated columns, one-hot encode categorical variables, and apply deviation encoding for high-cardinality factors. The code uses the `caret` package's `findCorrelation` for correlation-based feature removal, `model.matrix` for dummy encoding, and `doMC` for parallel correlation computation via `bigcorPar`. It also leverages `dplyr` for grouped summarization in deviation encoding, with the `ff` package for large correlation matrices.

Aaakansh9

49% match

Loading thumbnail…

Ranking Metrics

This example implements a collection of information retrieval ranking metrics in Python. It defines functions for mean reciprocal rank, R-precision, precision at k, average precision, mean average precision, and discounted cumulative gain, each accepting relevance scores in rank order. The code uses NumPy arrays to compute binary relevance and graded relevance calculations, with doctests illustrating expected outputs. The implementation follows standard definitions from academic resources on IR evaluation, such as precision-recall curves and learning-to-rank literature.

Aalvinthai

49% match

Loading thumbnail…

Hist : Comaraison Rf vs Xgb

This visualization compares feature importances from two machine learning models, Random Forest and XGBoost, using side-by-side histograms rendered as SVG bar charts with d3.v4. Each bar corresponds to a feature (e.g., Latitude, Beat, Description) and its height reflects the model’s importance score. Hovering over a bar in either chart highlights the matching bar in the other chart and displays a tooltip with the feature name and value. The code uses d3.scaleLinear for positioning, d3 selections for bar creation, and mouseover/mouseout events with transitions for interaction.

AAminaSGitHub

49% match

Loading thumbnail…

Plot mean expression of a gene vs other parameters and highlight specific genes using ggrepel

This example uses the scater package to calculate quality metrics for single-cell RNA-seq data, then plots mean gene expression against percent dropout and percent total expression. The ggplot2 code highlights a user-defined set of genes of interest in red, while ggrepel’s `geom_text_repel` labels those genes to avoid overlapping points. Two separate scatterplots are generated from the same `gene_info` data frame, each saved as an SVG, showing how specific genes relate to global expression distributions. The analysis relies on the `newSCESet` and `calculateQCMetrics` functions from scater to derive the plotted metrics.

FFloWuenne

48% match

Loading thumbnail…

Hist : Comaraison Rf vs Xgb

Two bar charts compare feature importance values for Random Forest and XGBoost models across 17 variables, with each chart labeled in French. Hovering over a bar in one chart highlights the corresponding bar in the other and displays a tooltip with the variable name and importance value. The visualization uses d3.v4 with SVG rendering, d3.scaleLinear for axes, and mouseover/mouseout events with transitions to animate the highlight effect. The data arrays are hardcoded in the script.

AAminaSGitHub

47% match

Loading thumbnail…

tut8

This example shows a machine-learning workflow for predicting Brexit referendum outcomes, comparing predicted versus actual Leave votes in a test dataset. It uses R to fit a final classification model, then applies the `predict` function to generate class predictions on held-out test data. The visualization or summary aspect centers on computing `base_accuracy`, the mean agreement between predicted and observed `Percent.Leave` values. The code relies on base R modeling and prediction APIs, with no plotting libraries, instead focusing on quantitative accuracy assessment rather than graphical output.

BBenHeubl

47% match