Result Prediction of Wordle Word Guessing Game Based on SVM And BP Neural Network Multi-Input-Multi-Output Regression Prediction Model

Abstract

Wordle is a popular daily puzzle game in which players can guess a five-letter word in six or fewer attempts. In order to predict the distribution column of user participation and guessing results, this paper established a support vector machine (SVM) regression model based on the historical data set of game results to predict the number of reports on March 1, 2023. Then, by extracting attribute features such as word frequency from words, the relationship between word attribute and report quantity under difficult mode is studied through multiple regression analysis and correlation coefficient analysis. Finally, a BP neural network multi-input-multi-output regression prediction model was established. The goodness of fit (R) of the model was 0.93671, and the distribution of correct guesses of the word "EERIE" was predicted. Through this regulation, the popularity of the game can be maintained to some extent, so that the number of people playing the game can maintain stability or growth.