Shulou(Shulou.com)06/03 Report--

This article mainly explains "what is the realization idea of the quantitative evaluation process". Interested friends may wish to take a look. The method introduced in this paper is simple, fast and practical. Next, let the editor take you to learn "what is the realization idea of the quantitative evaluation process"?

First, quantitative thinking

There are many complex businesses in the programming system that are difficult to understand, but it is necessary to do a quantitative analysis to give business personnel or operators, or users a reference standard, such as common index, sesame score, store grade, this kind of business evaluation criteria are very complex, because there are many factors that affect the results.

In the multi-dimensional business consideration model, there is a core concept called weight, which refers to the importance of a factor or index relative to something, which is different from the general proportion, reflecting not only the percentage of a factor or index, but also the relative importance of factors or indicators, which tends to contribute or importance. Typically, the weight of each dimension is between 0 and 1, and the sum of the weights of all dimensions is 1.

The concept of weight can be analyzed from an actual case, such as judging whether a customer is the object of key operations, usually in terms of login times per week, online duration, transaction volume and other dimensions. If customer An often logs in, but there is no core business transaction, customer B rarely logs in, but the business transaction is high, so the weight of login times here should be lower than the transaction volume dimension.

How to determine the proportion of weight, there are usually two ideas: one is to learn from the experience provided by professional business personnel and put it into the business to constantly try to optimize; second, according to the product analysis data, calculate the weight of each dimension, also need to constantly try to optimize in the business.

In fact, the quantification process of complex business scenarios is complex and long, and it is necessary to collect multi-dimensional data. Sometimes, not only periodic quantification is needed, such as the credit scores of several large companies, but also there may be scenarios of real-time analysis. Fraud risk control in financial business, there are also two scenarios integrated real-time recommendation system, which will use the quantization process.

Second, scene case

1. Comprehensive evaluation

Make a comprehensive evaluation of users, stores, products and other scenarios, abstract and analyze a complex thing through multiple dimensions, and generate simple and easy-to-understand evaluation results, such as store grade, product score, user comprehensive index, etc., and then provide a reference basis for each use scenario. From the point of view of the results, it may be easy to understand, but the analysis process of obtaining the results is relatively complex. Some scenarios may require periodic execution of the evaluation model, and some scenarios may require real-time calculation. It is also possible that the combination of the two situations, that is, relying on periodic evaluation, but also need to refer to real-time calculation.

2. Scene recommendation

This scenario is relatively complex, for example, the user searches, but selects a series of exclusions or necessary conditions, which is very common in the functions of the search class. in processing, it is necessary to analyze the user's search conditions with the highest degree of match. but also to do the optimal ranking based on the search results, there are two stages of evaluation, the first stage matches the optimal search conditions, and the second stage makes the optimal ranking of the matching results. Give the search results that the user wants as much as possible.

3. Risk control score

In the financial field, this is a very common risk control model, that is, multi-dimensional statistics of users, do dimension scores and then add together, the higher the risk control score, the greater the risk of the user, and then prevent high-risk transactions.

4. Financial management index

This scenario is very common. In the APP of finance and financial management, you must go through an evaluation system to judge the user's risk tolerance, such as conservative, positive, etc., when the product purchased by the user is of high risk, it will prompt the user to reevaluate it.

Third, the realization idea 1, the dimension rule table

Maintain a table of evaluation rules for a dimension. Classify_sign is understood as a division identity in the same business scenario, and weight identifies the proportion of the dimension in the evaluation.

CREATE TABLE `ID', int (11) NOT NULL AUTO_INCREMENT COMMENT 'primary key ID', `classify_ sign` varchar (50) COLLATE utf8mb4_unicode_ci DEFAULT' 'COMMENT' categorization ID', `rule_ value`varchar 'COLLATE utf8mb4_unicode_ci DEFAULT' 'COMMENT' rule description', `rule_ type`int (1) DEFAULT NULL COMMENT 'rule type: 1 accurate match, 2 range 3 fuzzy', `weight` decimal (10Jing 2) DEFAULT '0.00' COMMENT' weight distribution', PRIMARY KEY (`id`) ENGINE=InnoDB AUTO_INCREMENT=3 DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci COMMENT=' evaluation item rule' 2. Description rules

For the specific description of the rule, the core is two fields, the value of the rule and the result obtained by matching to the rule.

Public class RuleValue {/ * Rule value description * / private Object ruleValue; / * Rule matching result * / private Object ruleResult; / / Infrastructure public RuleValue (Object ruleValue, Object ruleResult) {this.ruleValue = ruleValue; this.ruleResult = ruleResult;} / / omitting Get and Set} 3, encapsulating matching values

In order to simplify the complexity of parameter transmission in the model, uniformly encapsulate the data of matching factors in a data model, here the two factors of city and label are used for process testing.

Public class MatchItem {/ / City private String city; / / tag private String tag; / / Infrastructure public MatchItem (String city, String tag) {this.city = city; this.tag = tag;} / / omitting Get and Set} 4, evaluation logic implementation

Here is only a simple description of the two cases, in the actual development scenario, the data and matching specifications are very complex, in the whole evaluation model implementation process needs to be constantly optimized.

@ Servicepublic class AssessBizService {private static Logger LOG = LoggerFactory.getLogger (AssessBizService.class); @ Resource private EvaluateRuleDao evaluateRuleDao; / * Business Evaluation process * / public void assessBiz (MatchItem matchItem) {/ / precise matching City EvaluateRuleEntity evaluateRule01 = evaluateRuleDao.getBySign ("assess-biz", 1); List cityRuleList = JSONArray.parseArray (evaluateRule01.getRuleValue (), RuleValue.class) For (RuleValue cityRule:cityRuleList) {if (cityRule.getRuleValue (). Equals (matchItem.getCity () {int result = Integer.parseInt (String.valueOf (cityRule.getRuleResult (); LOG.info ("match: {}, match result: {}", matchItem.getCity (), result*evaluateRule01.getWeight ()); break }} / / Fuzzy matching tag EvaluateRuleEntity evaluateRule02 = evaluateRuleDao.getBySign ("assess-biz", 3); List tagRuleList = JSONArray.parseArray (evaluateRule02.getRuleValue (), RuleValue.class) For (RuleValue tagRule:tagRuleList) {if (String.valueOf (tagRule.getRuleValue ()) .matches (matchItem.getTag () {int result = Integer.parseInt (String.valueOf (tagRule.getRuleResult (); LOG.info ("matches: {}, matching results: {}", matchItem.getTag (), result*evaluateRule02.getWeight ()); break Fourth, the source code address GitHub address https://github.com/cicadasmile/data-manage-parentGitEE address https://gitee.com/cicadasmile/data-manage-parent, I believe you have a deeper understanding of the "quantitative evaluation process is what the implementation of ideas", might as well to the actual operation of it! Here is the website, more related content can enter the relevant channels to inquire, follow us, continue to learn!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.