xgboost模型部署及注意相关事宜。

前段时间做的xgboost要部署成service，中途碰到不少坑，首先你部署的机器必须先装gcc环境，而且要6.0x版本

以上，4.x的版本的并不适应，看下代码service项目吧

首先是接口:

package com.dianping.text.classify.api.service;
public interface PredictionService {
	
   public double prediction(String text);
   public double prediction(String text,String cata);
   public double prediction(String  text,double rate);
   public double prediction(String text,double rate,int textlength);
   public double prediction(String text,String cata,double rate);
   public double prediction(String text,String cataS,double rate,int textlength);
}

实现接口类:

package com.dianping.text.classify.serviceimpl;
import java.io.BufferedReader;
import java.io.FileInputStream;
import java.io.IOException;
import java.io.InputStreamReader;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import com.alibaba.fastjson.JSONObject;
import com.dianping.text.classify.api.service.PredictionService;
import com.dianping.text.classify.util.DataLoader;
import com.dianping.text.classify.util.PathUtil;
import com.dianping.text.classify.util.Terms;
import com.meituan.nlp.util.WordUtil;
import ml.dmlc.xgboost4j.java.Booster;
import ml.dmlc.xgboost4j.java.XGBoost;
import ml.dmlc.xgboost4j.java.XGBoostError;
import org.ansj.splitWord.analysis.ToAnalysis;
import org.apache.commons.lang.StringUtils;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
public class ServiceImpl implements PredictionService {
	private static final Logger logger = LoggerFactory.getLogger(ServiceImpl.class);
	private Booster booster;
	private Map mapTerms ;
	private void init(){
	
		loadmapterms(this.getClass().getResource("/").getPath()+"model/mapterms.model");
		loadXgbosost(this.getClass().getResource("/").getPath()+"model/xgboost.model");
	
	}
	@Override
	public double predictionClassify(String text) {
		double result = DataLoader.getClassification(text, mapTerms, booster);
		return result;
	}
	@Override
	public List segmentation(String text) {
		if (StringUtils.isBlank(text)) {
			return null;
		}
		List list = new ArrayList();
		List lists = ToAnalysis.parse(text).getTerms();
		for (org.ansj.domain.Term str : lists) {
			if (!"w".equalsIgnoreCase(str.getNatureStr())) {
				list.add(str.getName());
			}
		}
		return list;
	}
	@Override
	public boolean isRepeat(String text, double rate, int contentLegth) {
		return WordUtil.isRepeat(text, rate, contentLegth);
	}
	private void loadXgbosost(String path) {
		
			try {
				booster = XGBoost.loadModel(path);
			} catch (XGBoostError e) {
				logger.error("loadd model error:", e);
			}
		
	}
	private void loadmapterms(String file) {
			mapTerms = new HashMap<>();
			BufferedReader in = null;
			String json = null;
			try {
				in = new BufferedReader(new InputStreamReader(
						new FileInputStream(file)));
				String line = in.readLine();
				while (line != null) {
					String[] lines = line.split("\t");
					Terms terms = JSONObject.parseObject(lines[1], Terms.class);
					mapTerms.put(lines[0], terms);
					line = in.readLine();
				}
			} catch (Exception e) {
				logger.error("load map error:", e);
				e.printStackTrace();
			} finally {
				if (in != null) {
					try {
						in.close();
					} catch (IOException e) {
						logger.error("close error:", e);
					}
				}
			}
		
	}
}

spring配置:

web.xml就不写了，总体来说，模型部署的时候一定要注意gcc环境的安装，否则部署会报错，还有mac 下编译的xgboost包并不适合在centos、乌班图上适应

否则也会报错，在centos上或者乌班图上部署一定要重新编译xgboost，把编译好的jar包maven到你们自己公司的仓库中去。