用C++写一个朴素贝叶斯算法

好的，以下是一个简单的朴素贝叶斯算法的C++实现，用于文本分类：

#include <iostream>
#include <fstream>
#include <vector>
#include <map>
#include <string>
#include <cmath>

using namespace std;

// 停用词
set<string> stopWords = {"the", "and", "a", "an", "in", "is", "it", "of", "to"};

// 分词函数
vector<string> tokenize(const string& text) {
    vector<string> tokens;
    string token;
    for (char c : text) {
        if (isalpha(c)) {
            token += tolower(c);
        } else {
            if (!token.empty()) {
                if (stopWords.count(token) == 0) {
                    tokens.push_back(token);
                }
                token.clear();
            }
        }
    }
    if (!token.empty()) {
        if (stopWords.count(token) == 0) {
            tokens.push_back(token);
        }
    }
    return tokens;
}

// 计算每个词的出现次数
map<string, int> wordCounts(const vector<string>& tokens) {
    map<string, int> counts;
    for (const string& token : tokens) {
        counts[token] += 1;
    }
    return counts;
}

// 训练函数
void train(const string& filename, map<string, map<string, int>>& wordCountsByClass, map<string, int>& docCountsByClass) {
    ifstream infile(filename);
    if (!infile) {
        cerr << "Error: could not open file " << filename << endl;
        exit(1);
    }
    string line;
    while (getline(infile, line)) {
        vector<string> tokens = tokenize(line);
        if (tokens.empty()) {
            continue;
        }
        string cls = tokens[0];
        docCountsByClass[cls] += 1;
        tokens.erase(tokens.begin());
        map<string, int>& counts = wordCountsByClass[cls];
        for (const pair<string, int>& p : wordCounts(tokens)) {
            counts[p.first] += p.second;
        }
    }
}

// 预测函数
string predict(const string& text, const map<string, map<string, int>>& wordCountsByClass, const map<string, int>& docCountsByClass) {
    vector<string> tokens = tokenize(text);
    if (tokens.empty()) {
        return "unknown";
    }
    double maxLogProb = -numeric_limits<double>::infinity();
    string argMaxClass;
    for (const pair<string, int>& p : docCountsByClass) {
        const string& cls = p.first;
        int docsInClass = p.second;
        double logProb = log(docsInClass) - log(docCountsByClass.size());
        const map<string, int>& counts = wordCountsByClass.at(cls);
        for (const string& token : tokens) {
            int tokenCount = counts.count(token) == 0 ? 0 : counts.at(token);
            logProb += log(tokenCount + 1) - log(counts.size() + 1);
        }
        if (logProb > maxLogProb) {
            maxLogProb = logProb;
            argMaxClass = cls;
        }
    }
    return argMaxClass;
}

int main() {
    // 训练集
    map<string, map<string, int>> wordCountsByClass;
    map<string, int> docCountsByClass;
    train("train.txt", wordCountsByClass, docCountsByClass);

    // 测试集
    ifstream testfile("test.txt");
    if (!testfile) {
        cerr << "Error: could not open file test.txt" << endl;
        exit(1);
    }
    string line;
    while (getline(testfile, line)) {
        string cls = predict(line, wordCountsByClass, docCountsByClass);
        cout << line << " => " << cls << endl;
    }

    return 0;
}

上述代码的训练数据和测试数据分别在train.txt和test.txt文件中。其中，训练数据的格式如下：

class1 sentence1
class1 sentence2
class2 sentence3
class2 sentence4
...

测试数据的格式与训练数据相同，只不过没有类别标签。代码中的分词函数和停用词可以根据实际情况进行调整。