Struct Booster

pub struct Booster { /* private fields */ }

Core model in LightGBM, containing functions for training, evaluating and predicting.

Implementations

impl Booster

pub fn from_file(filename: &str) -> Result<Self>

Load model from file.

pub fn from_string(model_description: &str) -> Result<Self>

Load model from string.

pub fn save_file(&self, filename: &str) -> Result<()>

Save model to file.

pub fn save_string(&self) -> Result<String>

Save model to string. This returns the same content that save_file writes into a file.

pub fn num_classes(&self) -> i32

Get the number of classes.

pub fn num_features(&self) -> i32

Get the number of features.

pub fn num_iterations(&self) -> i32

Get the number of iterations in the booster.

pub fn max_iterations(&self) -> i32

Get the maximum number of iterations used for prediction.

pub fn set_max_iterations(&mut self, max_iterations: i32) -> Result<()>

Sets the the maximum number of iterations for prediction.

pub fn train(dataset: Dataset, parameters: &Value) -> Result<Self>

Trains a new model using dataset and parameters.

Example

extern crate serde_json;
use lightgbm3::{Dataset, Booster};
use serde_json::json;

let xs = vec![vec![1.0, 0.1, 0.2, 0.1],
              vec![0.7, 0.4, 0.5, 0.1],
              vec![0.9, 0.8, 0.5, 0.1],
              vec![0.2, 0.2, 0.8, 0.7],
              vec![0.1, 0.7, 1.0, 0.9]];
let labels = vec![0.0, 0.0, 0.0, 1.0, 1.0];
let dataset = Dataset::from_vec_of_vec(xs, labels, true).unwrap();
let params = json!{
   {
        "num_iterations": 3,
        "objective": "binary",
        "metric": "auc"
    }
};
let bst = Booster::train(dataset, &params).unwrap();

Full set of parameters can be found on the official LightGBM docs: https://lightgbm.readthedocs.io/en/latest/Parameters.html

pub fn predict<T: DType>( &self, flat_x: &[T], n_features: i32, is_row_major: bool, ) -> Result<Vec<f64>>

Get predictions given &[f32] or &[f64] slice of features. The resulting vector will have the size of n_rows by n_classes.

pub fn predict_with_params<T: DType>( &self, flat_x: &[T], n_features: i32, is_row_major: bool, params: &str, ) -> Result<Vec<f64>>

Get predictions given &[f32] or &[f64] slice of features. The resulting vector will have the size of n_rows by n_classes.

Example:

use serde_json::json;
let y_pred = bst.predict_with_params(&xs, 10, true, "num_threads=1").unwrap();

pub fn raw_scores<T: DType>( &self, flat_x: &[T], n_features: i32, is_row_major: bool, ) -> Result<Vec<f64>>

Get raw scores given &[f32] or &[f64] slice of features. The resulting vector will have the size of n_rows by n_classes.

pub fn raw_scores_with_params<T: DType>( &self, flat_x: &[T], n_features: i32, is_row_major: bool, parameters: &str, ) -> Result<Vec<f64>>

Get raw scores given &[f32] or &[f64] slice of features. The resulting vector will have the size of n_rows by n_classes.

Example:

use serde_json::json;
let y_pred = bst.predict_with_params(&xs, 10, true, "num_threads=1").unwrap();

pub fn predict_from_vec_of_vec<T: DType>( &self, x: Vec<Vec<T>>, is_row_major: bool, ) -> Result<Vec<Vec<f64>>>

Predicts results for the given x and returns a vector or vectors (inner vectors will contain probabilities of classes per row). For regression the resulting inner vectors will have single element, so consider using predict method instead.

Input data example

let data = vec![vec![1.0, 0.1],
                vec![0.7, 0.4],
                vec![0.1, 0.7],
                vec![0.2, 0.5]];

Output data example for 3 classes:

let output = vec![vec![0.1, 0.8, 0.1],
                  vec![0.7, 0.2, 0.1],
                  vec![0.5, 0.4, 0.1],
                  vec![0.2, 0.2, 0.6],
];

pub fn feature_name(&self) -> Result<Vec<String>>

Gets features names.

pub fn feature_importance( &self, importance_type: ImportanceType, ) -> Result<Vec<f64>>

Get feature importance. Refer to ImportanceType

Trait Implementations

impl Drop for Booster

fn drop(&mut self)

Executes the destructor for this type.