Skip to content
Niko Aarnio edited this page Sep 20, 2024 · 15 revisions

EIS Toolkit implementation status

Status symbols:

  • 🟩 = Completed
  • 🟦 = In process / in review
  • πŸŸ₯ = Not yet started

For a more detailed set of included tools / functions and their documentation, check the documentation site.

Raster processing

Tool Status
Reproject raster 🟩
Resample raster 🟩
Clip raster 🟩
Snap raster 🟩
Window raster 🟩
Unify rasters 🟩
Surface derivative tools 🟩
Extract values from raster 🟩
Reclassify 🟩
Create constant raster 🟩
Unique conditions grid 🟩
Smoothing/filter tools 🟩
Distance to anomaly 🟩

Vector processing

Tool Status
Reproject vector 🟩
Interpolation - IDW 🟩
Interpolation - kriging 🟩
Density computation 🟩
Distance computation 🟩
Calculate geometry 🟩
Extract shared lines 🟩
Cell-based association 🟩
Rasterize 🟩
Proximity computation 🟦

Exploratory analyses

Tool Status
PCA 🟩
K-means clustering 🟩
DBSCAN 🟩
Descriptive statistics 🟩
Statistical hypothesis tests 🟩
Basic plots (scatter, line, bar, hist, kde, ecdf, heatmap, pairplot, regrplot) 🟩
Parallel coordinates plot 🟩
Feature importance 🟩
Local Moran's I 🟩

Data transformations

Tool Status
Binarize 🟩
Winsorize 🟩
Logarithmic 🟩
Min-max 🟩
Z-score normalize 🟩
Clip transform 🟩
One-hot encode 🟩
CoDa transformations 🟩

Training data tools

Tool Status
Balance data (SMOTETomek) 🟩
Split data 🟩
Data sampler 🟦

Prediction

Tool Status
Fuzzy overlay 🟩
Weights of evidence 🟩
Logistic regression 🟩
Random forest (regr+cls) 🟩
Gradient boosting (regr+csl) 🟩
MLP (multilayer perceptron, regr+cls) 🟩
BNN 🟦

Evaluation

Tool Status
Calculate base metrics 🟩
Summarize label metrics for binary classifier 🟩
Summarize probability metrics for binary classifier 🟩
Plot ROC curve 🟩
Plot DET curve 🟩
Plot precision-recall curve 🟩
Plot calibration curve 🟩
Plot distribution of predicted probabilities 🟩
Plot confusion matrix 🟩
Plot neural network training accuracy 🟩
Plot neural network training loss 🟩
Plot prediction area curves 🟩
Plot rate curve 🟩
Score model (accuracy, precision, recall, F1, MAE, MSE, RMSE, R2) 🟩