Java對List列表作均勻抽樣算法

從給定的list中作數據抽樣,須要保證採樣結果數據的分佈平衡。 

/**
     * 從min - max之間取出總數爲items的隨機數
     * @param min
     * @param items
     * @param max
     * @return
     */
    private static List<Integer> getRandomId(int min,int items,int max){
        List<Integer> ids = Lists.newArrayList();
        while(ids.size()<items){
            int randomId = ThreadLocalRandom.current().nextInt(max)+min;
            if(!ids.contains(randomId))ids.add(randomId);
        }
        ids.sort((x,y)->x-y);
        return ids;
    }

    /**
     * 從給定的list中作數據抽樣,須要保證採樣數據的均勻分佈
     * 步驟:
     * 1:將原數據分 m 份,
     * 2:取得每份的最小index和最大index
     * 3:從最小index到最大index之間取 n 個list id
     * 4:從原list中取出對應的id數據
     * @param srcDatas 源數據
     * @param sampleTotal 抽樣的數據總數
     * @param splitCopies 將原數據拆分的份數
     */
    public static List<String> sampleData(List<String> srcDatas,int sampleTotal,int splitCopies){
        if(splitCopies<=0)splitCopies = 1;
        int items = sampleTotal/splitCopies; //從每份中抽取的數據,數據總數將等於sampleTotal
        List<String> filterRes = Lists.newArrayList();//用於保存最終的抽樣結果
        ListSplit<String> listSplit = new ListSplit<>();//對list作拆分算法,源碼:https://my.oschina.net/u/2391658/blog/703032

        List<List<String>> splitRes = listSplit.split(srcDatas,splitCopies);

        int preListSize = 0;//初始化第一份List的最小下標
        for(int i=0;i<splitCopies;i++){
            List<String> listBlock = splitRes.get(i);//取出拆分後的list單元
            System.out.println(preListSize+"-->"+(preListSize+listBlock.size()-1));
            List<Integer> ids = getRandomId(preListSize,items,listBlock.size());//取到排序後的抽樣數據id

            System.out.println(Arrays.toString(ids.toArray()));
            ids.forEach(id->filterRes.add(srcDatas.get(id)));//取出list 下標id對應的值
            preListSize = preListSize+listBlock.size(); //從新初始化list的最小下標
        }
        System.out.println("抽樣結果:");
        System.out.println(Arrays.toString(filterRes.toArray()));
        return filterRes;
    }

    public static void main(String args[]){
        List<String> datas = Lists.newArrayList();
        for(int y=0;y<1004;y++)datas.add(y+"");//構造數據
        sampleData(datas,100,10);
    }

執行結果： 

0-->99
[5, 6, 25, 28, 29, 38, 48, 69, 72, 81]  //從每份list中抽樣結果
100-->199
[104, 142, 144, 145, 159, 164, 172, 174, 180, 188]
200-->299
[207, 212, 219, 228, 239, 250, 264, 281, 298, 299]
300-->399
[309, 313, 316, 324, 329, 331, 363, 364, 368, 377]
400-->499
[401, 417, 429, 441, 442, 448, 453, 484, 490, 493]
500-->599
[527, 532, 537, 544, 555, 556, 575, 584, 593, 596]
600-->699
[601, 628, 649, 655, 656, 659, 662, 675, 684, 696]
700-->799
[707, 709, 729, 734, 752, 763, 767, 770, 773, 774]
800-->899
[804, 820, 828, 831, 837, 848, 858, 865, 887, 893]
900-->999
[914, 926, 929, 940, 943, 954, 964, 979, 981, 998]
抽樣結果:
[5, 6, 25, 28, 29, 38, 48, 69, 72, 81, 104, 142, 144, 145, 159, 164, 172, 174, 180, 188, 207, 212, 219, 228, 239, 250, 264, 281, 298, 299, 309, 313, 316, 324, 329, 331, 363, 364, 368, 377, 401, 417, 429, 441, 442, 448, 453, 484, 490, 493, 527, 532, 537, 544, 555, 556, 575, 584, 593, 596, 601, 628, 649, 655, 656, 659, 662, 675, 684, 696, 707, 709, 729, 734, 752, 763, 767, 770, 773, 774, 804, 820, 828, 831, 837, 848, 858, 865, 887, 893, 914, 926, 929, 940, 943, 954, 964, 979, 981, 998]