PermutationFeatureImportance(
this RankingContext ctx,
IPredictionTransformer <IPredictor> model,
IDataView data,
string label = DefaultColumnNames.Label,
string groupId = DefaultColumnNames.GroupId,
string features = DefaultColumnNames.Features,
bool useFeatureWeightFilter = false,
int?topExamples = null)
{
return(PermutationFeatureImportance <RankerMetrics> .GetImportanceMetricsMatrix(
CatalogUtils.GetEnvironment(ctx),
model,
data,
idv => ctx.Evaluate(idv, label, groupId),
RankingDelta,
features,
useFeatureWeightFilter,
topExamples));
}
[ConditionalFact(typeof(Environment), nameof(Environment.Is64BitProcess))] // LightGBM is 64-bit only
public void LightGBMRanking()
{
var env = new MLContext(seed: 0);
var dataPath = GetDataPath(TestDatasets.adultRanking.trainFilename);
var dataSource = new MultiFileSource(dataPath);
var ctx = new RankingContext(env);
var reader = TextLoader.CreateReader(env,
c => (label: c.LoadFloat(0), features: c.LoadFloat(9, 14), groupId: c.LoadText(1)),
separator: '\t', hasHeader: true);
LightGbmRankingModelParameters pred = null;
var est = reader.MakeNewEstimator()
.Append(r => (r.label, r.features, groupId: r.groupId.ToKey()))
.Append(r => (r.label, r.groupId, score: ctx.Trainers.LightGbm(r.label, r.features, r.groupId, onFit: (p) => { pred = p; })));
var pipe = reader.Append(est);
Assert.Null(pred);
var model = pipe.Fit(dataSource);
Assert.NotNull(pred);
var data = model.Read(dataSource);
var metrics = ctx.Evaluate(data, r => r.label, r => r.groupId, r => r.score);
Assert.NotNull(metrics);
Assert.True(metrics.Ndcg.Length == metrics.Dcg.Length && metrics.Dcg.Length == 3);
Assert.InRange(metrics.Dcg[0], 1.4, 1.6);
Assert.InRange(metrics.Dcg[1], 1.4, 1.8);
Assert.InRange(metrics.Dcg[2], 1.4, 1.8);
Assert.InRange(metrics.Ndcg[0], 36.5, 37);
Assert.InRange(metrics.Ndcg[1], 36.5, 37);
Assert.InRange(metrics.Ndcg[2], 36.5, 37);
}