treeboost 0.1.0

High-performance Gradient Boosted Decision Tree engine for large-scale tabular data
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278

//! Integration tests for data pipeline

use polars::prelude::*;
use treeboost::booster::{GBDTConfig, GBDTModel};
use treeboost::dataset::{DataPipeline, PipelineConfig};

#[test]
fn test_data_pipeline_end_to_end() {
    // Create a DataFrame with mixed types including categoricals
    let df = df! {
        "price" => &[100.0, 150.0, 200.0, 180.0, 220.0, 90.0, 300.0, 250.0, 175.0, 400.0,
                     120.0, 160.0, 210.0, 190.0, 230.0, 95.0, 310.0, 260.0, 185.0, 420.0],
        "sqft" => &[1000.0, 1200.0, 1500.0, 1400.0, 1600.0, 900.0, 2000.0, 1800.0, 1350.0, 2500.0,
                    1050.0, 1250.0, 1550.0, 1450.0, 1650.0, 950.0, 2050.0, 1850.0, 1400.0, 2600.0],
        "bedrooms" => &[2.0, 3.0, 3.0, 3.0, 4.0, 2.0, 4.0, 4.0, 3.0, 5.0,
                        2.0, 3.0, 3.0, 3.0, 4.0, 2.0, 4.0, 4.0, 3.0, 5.0],
        "neighborhood" => &["downtown", "suburbs", "downtown", "suburbs", "downtown",
                            "rural", "downtown", "suburbs", "downtown", "suburbs",
                            "downtown", "suburbs", "downtown", "suburbs", "downtown",
                            "rural", "downtown", "suburbs", "downtown", "typo_rare"],
        "target" => &[250.0, 280.0, 350.0, 320.0, 400.0, 180.0, 500.0, 450.0, 300.0, 600.0,
                      260.0, 290.0, 360.0, 330.0, 410.0, 185.0, 510.0, 460.0, 310.0, 550.0]
    }
    .unwrap();

    // Create pipeline with configuration
    let pipeline = DataPipeline::new(
        PipelineConfig::new()
            .with_num_bins(16)
            .with_cms_params(0.01, 0.99, 2) // min_count=2 to filter "typo_rare" and "rural"
            .with_smoothing(5.0),
    );

    // Process for training
    let (dataset, state, _filtered_df) = pipeline
        .process_for_training(df.clone(), "target", Some(&["neighborhood"]))
        .expect("Pipeline should succeed");

    assert_eq!(dataset.num_rows(), 20);
    assert_eq!(dataset.num_features(), 4); // price, sqft, bedrooms, neighborhood

    // Check that categorical encoding state was learned
    assert_eq!(state.categorical_encodings.len(), 1);
    assert_eq!(state.categorical_encodings[0].name, "neighborhood");

    // Train a model
    let config = GBDTConfig::new()
        .with_num_rounds(20)
        .with_max_depth(4)
        .with_learning_rate(0.1);

    let model = GBDTModel::train_binned(&dataset, config).expect("Training should succeed");

    // Make predictions
    let predictions = model.predict(&dataset);
    assert_eq!(predictions.len(), 20);

    // Calculate MSE - should be reasonable
    let targets = dataset.targets();
    let mse: f32 = predictions
        .iter()
        .zip(targets.iter())
        .map(|(p, t): (&f32, &f32)| (p - t).powi(2))
        .sum::<f32>()
        / predictions.len() as f32;

    assert!(mse < 10000.0, "MSE {} is too high for training data", mse);

    // Test inference with new data (including unseen category)
    let test_df = df! {
        "price" => &[175.0, 275.0],
        "sqft" => &[1300.0, 1900.0],
        "bedrooms" => &[3.0, 4.0],
        "neighborhood" => &["downtown", "unseen_area"]  // includes unseen category
    }
    .unwrap();

    let (_test_preprocessed_df, test_dataset) = pipeline
        .process_for_inference(test_df, &state)
        .expect("Inference pipeline should succeed");

    assert_eq!(test_dataset.num_rows(), 2);

    let test_predictions = model.predict(&test_dataset);
    assert_eq!(test_predictions.len(), 2);

    // Predictions should be positive and reasonable
    for pred in &test_predictions {
        assert!(*pred > 0.0, "Prediction should be positive");
        assert!(*pred < 1000.0, "Prediction should be reasonable");
    }
}

#[test]
fn test_pipeline_rare_category_filtering() {
    // Create data with rare categories that should be filtered
    let df = df! {
        "feature" => &[1.0; 100],
        "category" => {
            let mut cats = vec!["frequent".to_string(); 50];
            cats.extend(vec!["also_frequent".to_string(); 30]);
            cats.extend(vec!["rare1".to_string(); 5]);
            cats.extend(vec!["rare2".to_string(); 5]);
            cats.extend(vec!["very_rare1".to_string(); 3]);
            cats.extend(vec!["very_rare2".to_string(); 3]);
            cats.extend(vec!["typo1".to_string(); 2]);
            cats.extend(vec!["typo2".to_string(); 2]);
            cats
        },
        "target" => &(0..100).map(|i| i as f64).collect::<Vec<_>>()
    }
    .unwrap();

    let pipeline = DataPipeline::new(
        PipelineConfig::new()
            .with_num_bins(8)
            .with_cms_params(0.01, 0.99, 10) // min_count=10 to filter many categories
            .with_smoothing(1.0),
    );

    let (_dataset, state, _filtered_df) = pipeline
        .process_for_training(df, "target", Some(&["category"]))
        .expect("Pipeline should succeed");

    // Should have filtered rare categories
    let cat_state = &state.categorical_encodings[0];

    // Only "frequent" (50) and "also_frequent" (30) should remain
    // All others have count < 10
    assert_eq!(
        cat_state.category_mapping.category_to_idx.len(),
        2,
        "Expected 2 frequent categories, got {}",
        cat_state.category_mapping.category_to_idx.len()
    );

    // Verify the frequent ones are kept
    let kept_cats: Vec<&str> = cat_state
        .category_mapping
        .category_to_idx
        .iter()
        .map(|(name, _)| name.as_str())
        .collect();

    assert!(kept_cats.contains(&"frequent"), "Should keep 'frequent'");
    assert!(
        kept_cats.contains(&"also_frequent"),
        "Should keep 'also_frequent'"
    );
}

#[test]
fn test_pipeline_target_encoding_prevents_leakage() {
    // Create data where target encoding with leakage would be obvious
    let df = df! {
        "category" => &["A", "A", "A", "B", "B", "B", "C", "C", "C", "C"],
        "target" => &[10.0, 20.0, 30.0, 100.0, 200.0, 300.0, 1.0, 2.0, 3.0, 4.0]
    }
    .unwrap();

    let pipeline = DataPipeline::new(
        PipelineConfig::new()
            .with_num_bins(8)
            .with_cms_params(0.01, 0.99, 1) // Keep all categories
            .with_smoothing(0.0), // No smoothing for clearer test
    );

    let (_dataset, state, _filtered_df) = pipeline
        .process_for_training(df, "target", Some(&["category"]))
        .expect("Pipeline should succeed");

    // With ordered encoding:
    // - First "A" (row 0) sees no prior data -> encoded as 0
    // - First "B" (row 3) sees global mean of rows 0-2 = (10+20+30)/3 = 20
    // - If there was leakage, B would be encoded as mean(100,200,300)=200

    // The key property: each row's encoding doesn't include its own target
    // We verify this indirectly by checking the encoding map
    let cat_state = &state.categorical_encodings[0];

    // The final encoding map uses ALL data, but the training used ordered encoding
    // A's final mean: (10+20+30)/3 = 20
    // B's final mean: (100+200+300)/3 = 200
    // C's final mean: (1+2+3+4)/4 = 2.5

    let enc_a = cat_state.encoding_map.encode("A");
    let enc_b = cat_state.encoding_map.encode("B");
    let enc_c = cat_state.encoding_map.encode("C");

    // A should be encoded ~20, B ~200, C ~2.5
    assert!(
        enc_a < enc_b,
        "A (low target) should have lower encoding than B (high target)"
    );
    assert!(
        enc_c < enc_a,
        "C (very low target) should have lowest encoding"
    );
}

#[test]
fn test_raw_prediction_equivalence() {
    // Create a dataset using the data pipeline (which sets up proper bin_boundaries)
    let df = df! {
        "f0" => &[1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0,
                  11.0, 12.0, 13.0, 14.0, 15.0, 16.0, 17.0, 18.0, 19.0, 20.0,
                  21.0, 22.0, 23.0, 24.0, 25.0, 26.0, 27.0, 28.0, 29.0, 30.0,
                  31.0, 32.0, 33.0, 34.0, 35.0, 36.0, 37.0, 38.0, 39.0, 40.0,
                  41.0, 42.0, 43.0, 44.0, 45.0, 46.0, 47.0, 48.0, 49.0, 50.0],
        "f1" => &[50.0, 49.0, 48.0, 47.0, 46.0, 45.0, 44.0, 43.0, 42.0, 41.0,
                  40.0, 39.0, 38.0, 37.0, 36.0, 35.0, 34.0, 33.0, 32.0, 31.0,
                  30.0, 29.0, 28.0, 27.0, 26.0, 25.0, 24.0, 23.0, 22.0, 21.0,
                  20.0, 19.0, 18.0, 17.0, 16.0, 15.0, 14.0, 13.0, 12.0, 11.0,
                  10.0, 9.0, 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0],
        "f2" => &[5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0,
                  5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0,
                  5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0,
                  5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0,
                  5.0, 10.0, 15.0, 20.0, 25.0, 30.0, 35.0, 40.0, 45.0, 50.0],
        "target" => &[10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0,
                      110.0, 120.0, 130.0, 140.0, 150.0, 160.0, 170.0, 180.0, 190.0, 200.0,
                      210.0, 220.0, 230.0, 240.0, 250.0, 260.0, 270.0, 280.0, 290.0, 300.0,
                      310.0, 320.0, 330.0, 340.0, 350.0, 360.0, 370.0, 380.0, 390.0, 400.0,
                      410.0, 420.0, 430.0, 440.0, 450.0, 460.0, 470.0, 480.0, 490.0, 500.0]
    }
    .unwrap();

    // Process with data pipeline to get proper bin boundaries
    let pipeline = DataPipeline::new(PipelineConfig::new().with_num_bins(16));
    let (dataset, _state, _filtered_df) = pipeline
        .process_for_training(df.clone(), "target", None)
        .expect("Pipeline should succeed");

    // Train a model
    let config = GBDTConfig::new()
        .with_num_rounds(20)
        .with_max_depth(4)
        .with_learning_rate(0.1);

    let model = GBDTModel::train_binned(&dataset, config).expect("Training should succeed");

    // Get predictions using binned data
    let binned_predictions = model.predict(&dataset);

    // Get raw feature values (row-major)
    let num_rows = 50;
    let num_features = 3;
    let mut raw_features = Vec::with_capacity(num_rows * num_features);

    // f0, f1, f2 for each row
    for i in 0..num_rows {
        raw_features.push((i + 1) as f64); // f0
        raw_features.push((50 - i) as f64); // f1
        raw_features.push(((i % 10) * 5 + 5) as f64); // f2
    }

    // Get predictions using raw values
    let raw_predictions = model.predict_raw(&raw_features);

    // Both should have the same length
    assert_eq!(binned_predictions.len(), raw_predictions.len());

    // Predictions should be very close (within floating-point tolerance)
    // Note: Due to binning discretization, there may be small differences
    let max_diff: f32 = binned_predictions
        .iter()
        .zip(raw_predictions.iter())
        .map(|(b, r)| (b - r).abs())
        .fold(0.0f32, f32::max);

    // The predictions should be reasonably close
    // (they may differ slightly due to binning boundary edge cases)
    assert!(
        max_diff < 50.0,
        "Max difference between binned and raw predictions: {} (expected < 50.0)",
        max_diff
    );
}