Disclosed in the present invention is a sampling simulation-based quick A/B testing method: A) randomly generating a testing parameter combination for each sample as an advertising parameter; B) using the advertising parameter in operation, and collecting statistics about LTV after an operation cycle; C) constructing a machine learning model; D) adjusting a sampling space and sampling interval, performing sampling again, and respectively plugging multiple advertising parameters into the machine learning model to obtain respective LTV prediction values; E) selecting the top M values to perform actual operation testing, determining whether LTV operation testing valves conform to the LTV prediction values, and if yes, considering the corresponding advertising parameters as the optimal parameters and proceeding to F), otherwise, returning to D); F) using the optimal parameters in actual production; and G) periodically performing A)-E), so that the optimal parameters change as the environment changes. According to the present invention, optimal parameters can be found with less A/B tests, and the A/B testing efficiency is improved.
展开▼
