package cn.edu.bjut.textmining.chapter5;

import cn.edu.bjut.textmining.chapter5.support.DocumentRegistry;
import cn.edu.bjut.textmining.chapter5.support.ResultWriter;
import cn.edu.bjut.textmining.chapter5.support.SampleNews;
import cn.edu.bjut.textmining.chapter5.support.TNewsCorpusLoader;
import cn.edu.bjut.textmining.chapter5.support.TextFeatures;

/** 对单位化TF-IDF进行球面K-Means。 */
public class TextKMeansDemo {
    public static void main(String[] args) {
        try {
            java.util.List<TNewsCorpusLoader.NewsRecord> records = SampleNews.load();
            TextFeatures.Result features = TextFeatures.fit(TNewsCorpusLoader.titles(records), false);
            double[][] vectors = features.vectors;
            KMeansEngine.Result result = KMeansEngine.fit(vectors, 3, 30, 0, 3, 6);
            System.out.printf(
                    "迭代轮数=%d，收敛=%s，目标值=%.9f%n",
                    result.iterations(),
                    result.converged(),
                    KMeansEngine.objective(vectors, result));
            SampleNews.printClusters(result.labels());
            ResultWriter.features("news_sample", features);
            DocumentRegistry.write("news_sample", records, features, result.labels(),
                    DocumentRegistry.SMALL_CONFIG);
            System.out.println("记录与向量行清单：outputs/chapter5/news_sample_document_registry.tsv");
        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}