Machine learning algorithms with dart

Following algorithms are implemented:

  • Linear regression:

    • gradient descent models (batch, mini-batch, stochastic) with ridge regularization
    • lasso model (feature selection model)
  • Linear classifier:

    • Logistic regression (with "one-vs-all" multinomial classification)


A simple usage example (Linear regression with stochastic gradient descent):

Import all necessary packages:

import 'dart:io';
import 'dart:async';
import 'dart:convert';
import 'package:ml_algo/ml_algo.dart';
import 'package:csv/csv.dart' as csv;

Read csv-file advertising.csv with test data:

final csvCodec = csv.CsvCodec(eol: '\n');
final input = File('example/datasets/advertising.csv').openRead();
final fields = (await input.transform(utf8.decoder)

Data in this file is represented by 200 lines, every line contains 4 elements. First 3 elements of every line are features and the last one is label. Let's extract features from the data. Declare utility method extractFeatures, that extracts 3 elements from every line:

List<double> extractFeatures(List<dynamic> item) => item.sublist(0, 3)
      .map((dynamic feature) => (feature as num).toDouble())

...and finally get all features:

final features = fields
  .toList(growable: false);

...and labels (last element of a every line)

final labels = Float32x4VectorFactory.from(<dynamic> item) => (item.last as num).toDouble()));

Create an instance of CrossValidator class for evaluating quality of our predictor

final validator = CrossValidator<Float32x4>.KFold();

Create a linear regressor instance with stochastic gradient descent optimizer:

final sgdRegressor = GradientRegressor(type: GradientType.stochastic, iterationLimit: 100000,
                         learningRate: 1e-5, learningRateType: LearningRateType.constant);

Evaluate our model via MAPE-metric:

final scoreMAPE = validator.evaluate(sgdRegressor, Float32x4Matrix.from(features), labels, metric: MetricType.mape);

Let's print score:

print("score (MAPE): ${scoreMAPE}");

We will see something like this:

score (MAPE): 31.221150755882263

For more examples please see examples folder