README.jmd

# JLBoostMLJ.jl

The [MLJ.jl](https://github.com/alan-turing-institute/MLJ.jl) interface to [JLBoost.jl](https://github.com/xiaodaigh/JLBoost.jl), a hackable implementation of Gradient Boosting Regression Trees.


## Usage Example

```julia
using RDatasets;
iris = dataset("datasets", "iris");
iris[!, :is_setosa] = iris.Species .== "setosa";

using MLJ, JLBoostMLJ;
X, y = unpack(iris, x->!(x in [:is_setosa, :Species]), ==(:is_setosa));

using JLBoostMLJ:JLBoostClassifier;
model = JLBoostClassifier()
```

### Using MLJ machines

Put the model and data in a machine

```julia
mljmachine  = machine(model, X, y)
```

Fit model using machine

```julia
fit!(mljmachine)
```

Predict using machine

```julia
predict(mljmachine, X)
```

Feature importance using machine

```julia
feature_importance(fitted_params(mljmachine).fitresult, X, y)
```

#### Hyperparameter tuning

Data preparation: need to convert `y` to categorical

```julia
y_cate = categorical(y)
```

Set up some hyperparameter ranges

```julia
using JLBoost, JLBoostMLJ, MLJ
jlb = JLBoostClassifier()
r1 = range(jlb, :nrounds, lower=1, upper = 6)
r2 = range(jlb, :max_depth, lower=1, upper = 6)
r3 = range(jlb, :eta, lower=0.1, upper=1.0)
```

Set up the machine
```julia
tm = TunedModel(model = jlb, ranges = [r1, r2, r3], measure = cross_entropy)
m = machine(tm, X, y_cate)
```

Fit it!
```julia
fit!(m)
```

Inspected the tuned parameters
```julia
fitted_params(m).best_model.max_depth
fitted_params(m).best_model.nrounds
fitted_params(m).best_model.eta
```

### Simple Fitting

Fit the model with `verbosity = 1`
```julia
mljmodel = fit(model, 1, X, y)
```
Predicting using the model

```julia
predict(model, mljmodel.fitresult, X)
```

Feature Importance for simple fitting
One can obtain the feature importance using the `feature_importance` function

```julia
feature_importance(mljmodel.fitresult.treemodel, X, y)
```