rcf3 0.1.0

A Rust implementation of the Random Cut Forest algorithm for anomaly detection.
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
279

// Test all code examples from README to ensure they compile and run correctly

#[cfg(test)]
mod readme_examples {
    use rcf3::Forest;

    #[test]
    fn test_creating_forest_basic() -> Result<(), Box<dyn std::error::Error>> {
        let forest = Forest::builder(2, 1) // 2D input, no shingling
            .num_trees(50)
            .capacity(256)
            .build()?;

        assert_eq!(forest.num_trees(), 50);
        Ok(())
    }

    #[test]
    fn test_creating_forest_with_time_series() -> Result<(), Box<dyn std::error::Error>> {
        let forest = Forest::builder(4, 8) // 4D input, window size 8
            .internal_shingling(true)
            .num_trees(100)
            .capacity(512)
            .time_decay(0.01)
            .build()?;

        assert_eq!(forest.num_trees(), 100);
        Ok(())
    }

    #[test]
    fn test_creating_forest_from_config() -> Result<(), Box<dyn std::error::Error>> {
        use rcf3::{Forest, RcfConfig};

        let config = RcfConfig::new(3)
            .with_num_trees(75)
            .with_capacity(512)
            .with_shingle_size(4);

        let forest = Forest::from_config(&config)?;
        assert_eq!(forest.num_trees(), 75);
        Ok(())
    }

    #[test]
    fn test_basic_operations() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(2, 1).capacity(256).num_trees(50).build()?;

        // Update the forest with a new observation
        let point = vec![1.5, 2.3];
        forest.update(&point)?;

        // Check if the forest has warmed up and get score
        if forest.is_ready() {
            let score = forest.score(&point)?;
            println!("Anomaly score: {}", score);
            assert!(score >= 0.0);
        }

        // Get the number of observations processed
        println!("Entries seen: {}", forest.entries_seen());

        Ok(())
    }

    #[test]
    fn test_scoring_methods() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(3, 1).capacity(256).num_trees(50).build()?;

        // Feed some data to warm up the forest
        for _ in 0..100 {
            forest.update(&vec![1.5, 2.3, -0.5])?;
        }

        let point = vec![1.5, 2.3, -0.5];

        // Anomaly Score (RCF Score)
        let score = forest.score(&point)?;
        assert!(score >= 0.0);
        println!("RCF Score: {}", score);

        // Displacement Score
        let displacement = forest.displacement_score(&point)?;
        assert!(displacement >= 0.0);
        println!("Displacement Score: {}", displacement);

        // Density Estimate
        let density = forest.density(&point)?;
        assert!(density >= 0.0);
        println!("Density: {}", density);

        Ok(())
    }

    #[test]
    fn test_feature_attribution() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(3, 1).capacity(256).num_trees(50).build()?;

        // Feed normal data first
        for _ in 0..100 {
            forest.update(&vec![1.0, 2.0, 3.0])?;
        }

        // Test with anomalous point
        let point = vec![1.5, 2.3, 100.0]; // Third dimension is anomalous
        let attribution = forest.attribution(&point)?;

        for (i, attr) in attribution.iter().enumerate() {
            println!(
                "Dimension {}: below={}, above={}",
                i, attr.below, attr.above
            );
            assert!(attr.below >= 0.0);
            assert!(attr.above >= 0.0);
        }

        Ok(())
    }

    #[test]
    fn test_neighborhood_search() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(2, 1).capacity(256).num_trees(50).build()?;

        // Feed some data points to build up the forest
        let data = vec![
            vec![1.0, 2.0],
            vec![1.1, 2.1],
            vec![1.2, 2.2],
            vec![1.3, 2.3],
            vec![1.4, 2.4],
            vec![5.0, 6.0],
            vec![5.1, 6.1],
            vec![5.2, 6.2],
        ];

        for point in &data {
            forest.update(&point)?;
        }

        let query_point = vec![1.5, 2.3];
        let neighbors = forest.near_neighbors(&query_point, 3, 50)?;

        println!("Found {} neighbors:", neighbors.len());
        for neighbor in neighbors {
            println!(
                "Distance: {}, Score: {}, Point: {:?}",
                neighbor.distance, neighbor.score, neighbor.point
            );
        }

        Ok(())
    }

    #[test]
    fn test_missing_value_imputation() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(3, 1).capacity(256).num_trees(50).build()?;

        // Feed some complete data to train
        for i in 0..100 {
            forest.update(&vec![1.0 + (i as f32) * 0.01, 2.0, 3.0])?;
        }

        // Test imputation with a missing value at index 1
        let point = vec![1.5, f32::NAN, 3.0];
        let missing = vec![1];
        let imputed = forest.impute(&point, &missing, 1.0)?;

        println!("Imputed value at index 1: {}", imputed[1]);
        assert!(!imputed[1].is_nan());
        assert!(imputed.len() == 3);

        Ok(())
    }

    #[test]
    #[cfg(feature = "serde")]
    fn test_serialization() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(2, 1).capacity(256).num_trees(50).build()?;

        // Feed some data
        for _ in 0..50 {
            forest.update(&vec![1.5, 2.3])?;
        }

        // Save to string
        let json_str = forest.to_json()?;
        assert!(!json_str.is_empty());
        println!("JSON length: {}", json_str.len());

        // Load from string
        let loaded = Forest::from_json(&json_str)?;
        assert_eq!(loaded.num_trees(), forest.num_trees());

        Ok(())
    }

    #[test]
    fn test_anomaly_detection_example() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(3, 1).capacity(256).num_trees(50).build()?;

        // Warm up the forest with many normal data points
        for i in 0..200 {
            let val = (i as f32) * 0.01;
            forest.update(&vec![1.0 + val, 2.0 + val, 3.0 + val])?;
        }

        let data = vec![
            vec![1.0, 2.0, 3.0],
            vec![1.1, 2.1, 3.1],
            vec![1.2, 2.2, 3.2],
            vec![100.0, 200.0, 300.0], // Extreme anomaly
            vec![1.3, 2.3, 3.3],
        ];

        let mut anomaly_count = 0;
        for point in data {
            // Online inference order: score first, then update.
            if forest.is_ready() {
                let score = forest.score(&point)?;
                let attribution = forest.attribution(&point)?;

                println!("Point: {:?}, Score: {}", point, score);

                // Lower threshold since we're detecting a very extreme anomaly
                if score > 0.1 {
                    println!("Anomaly detected: score={}", score);
                    for (i, attr) in attribution.iter().enumerate() {
                        println!("  Dimension {}: {:.2}", i, attr.above);
                    }
                    anomaly_count += 1;
                }
            }

            forest.update(&point)?;
        }

        println!("Total anomalies detected: {}", anomaly_count);
        assert!(anomaly_count > 0); // We expect to detect the anomaly

        Ok(())
    }

    #[test]
    fn test_time_series_forecasting() -> Result<(), Box<dyn std::error::Error>> {
        let mut forest = Forest::builder(4, 8)
            .internal_shingling(true)
            .capacity(512)
            .num_trees(50)
            .build()?;

        // Feed observations one at a time (time series)
        let stream = vec![
            vec![1.0, 2.0, 3.0, 4.0],
            vec![1.1, 2.1, 3.1, 4.1],
            vec![1.2, 2.2, 3.2, 4.2],
            vec![1.3, 2.3, 3.3, 4.3],
            vec![1.4, 2.4, 3.4, 4.4],
            vec![1.5, 2.5, 3.5, 4.5],
            vec![1.6, 2.6, 3.6, 4.6],
            vec![1.7, 2.7, 3.7, 4.7],
            vec![1.8, 2.8, 3.8, 4.8],
            vec![1.9, 2.9, 3.9, 4.9],
        ];

        for point in stream {
            forest.update(&point)?;
        }

        // Predict the next 5 observations
        if forest.is_ready() {
            let predictions = forest.extrapolate(5)?;
            println!("Predictions (flat list): {:?}", predictions);
            // Returns a flat list of length 5 * input_dim = 5 * 4 = 20
            assert_eq!(predictions.len(), 20);
        }

        Ok(())
    }
}