Java implementation of weighted stochastic algorithm

I. Overview

Usually, the weight of random algorithm, from the different weights of n elements randomly selected, and make the overall selection results are distributed according to the weight. such as ad delivery, load balancing and so on.

If there are 4 elements a, B, C, D, weights are 1, 2, 3, 4, the proportion of a:b:c:d in random results to be 1:2:3:4.

　　General idea : Add weights for each element a (1)-B (3)-C (6)-D (10), the weight of the 4 elements of the jurisdictional range is [0,1], [1,3], [3,6], [6,10]. Then randomly out a random number between [0,10]. The element that follows the interval is the element that is hit by weight.

　　Implementation Method :

Using TreeMap, a tree is constructed as follows:
B (3)
/      \
/         \
A (1) D (10)
/
/
C (6)

The target element can then be found using Treemap.tailmap (). Firstkey ().

Of course, you can also use array + binary lookup to implement.

Second, the source code

12345678910111213141516171819202122232425262728293031

packagecom.xxx.utils;importcom.google.common.base.Preconditions;importorg.apache.commons.math3.util.Pair;importorg.slf4j.Logger;importorg.slf4j.LoggerFactory;importjava.util.List;import java.util.SortedMap;importjava.util.TreeMap;publicclassWeightRandom<K,V extendsNumber> {    privateTreeMap<Double, K> weightMap = newTreeMap<Double, K>();    privatestaticfinalLogger logger = LoggerFactory.getLogger(WeightRandom.class);    publicWeightRandom(List<Pair<K, V>> list) {        Preconditions.checkNotNull(list, "list can NOT be null!");        for(Pair<K, V> pair : list) {            doublelastWeight = this.weightMap.size() == 0? 0: this.weightMap.lastKey().doubleValue();//统一转为double            this.weightMap.put(pair.getValue().doubleValue() + lastWeight, pair.getKey());//权重累加        }    }    publicK random() {        doublerandomWeight = this.weightMap.lastKey() * Math.random();        SortedMap<Double, K> tailMap = this.weightMap.tailMap(randomWeight, false);        returnthis.weightMap.get(tailMap.firstKey());    }}

　　

　　

Third, performance

4 elements A, B, C, D, their weights are 1, 2, 3, 4, run 100 million times, the results are as follows:

Elements	Hit count	Error rate
A	10004296	0.043%
B	19991132	0.0443%
C	30000882	0.0029%
D	40003690	0.0092%

From the results, it can be seen that the accuracy rate is above 99.95% .

Java implementation of weighted stochastic algorithm