This package provides algorithms for learning decision lists based on different logics. At the moment, it provides a starter for a CN2-like, sequential covering algorithm. The algorithm deploys a beam search to learn conjunctions of arbitrary formulae (including propositional atoms).
Simply type the following commands in Julia's REPL:
# Install packages
using Pkg; Pkg.add("MLJ");
using Pkg; Pkg.add("ModalDecisionLists");
# Import packages
using MLJ
using ModalDecisionLists
using Random
# Load an example dataset
X, y = MLJ.@load_iris()
N = length(y)
# Instantiate an MLJ machine
mach = machine(ExtendedSequentialCovering(), X, y)
# Split dataset
p = randperm(N)
train_idxs, test_idxs = p[1:round(Int, N*.8)], p[round(Int, N*.8)+1:end]
# Fit
fit!(mach, rows=train_idxs)
# Perform predictions, compute accuracy
yhat = predict_mode(mach, rows = test_idxs)
accuracy = MLJ.accuracy(yhat, y[test_idxs])
# Access & inspect model
dlist = fitted_params(mach).fitresult.model
printmodel(dlist; show_metrics = true, show_subtree_metrics=true)
# Make test instances flow into the model
test_dlist = deepcopy(dlist)
apply!(test_dlist, slicedataset(PropositionalLogiset(X), test_idxs), y[test_idxs])
# Extract rules that perform well in test
interestingrules = listrules(test_dlist; min_lift = 1.0, min_coverage = 0.05, normalize = true)
printmodel.(interestingrules; show_metrics = (; round_digits = 2));ModalDecisionLists.jl lives within the Sole.jl framework for symbolic machine learning.