Prediction and Analysis of Non-linear Data based on BP Neural Network

Wordle is a world-class and challenging game, with the goal of guessing hidden target words under specific constraints. For better development of Wordle game, the prediction and analysis of future Wordle player data has become crucial. The traditional analysis method is to build multiple linear or nonlinear regression models to predict the popularity of games. The performance data of Wordle players can be transformed into a non-linear data that contains complex mathematical rules. Base on the non-linear data, we adopted a efficient grey model (GM) to predict the number of future players of Wordle game, and utilized a effective neural network model (NNM) to fit and predict the relevant percentage that all players can guess 1/2/3/4/5/6/X times for a hidden target word in Wordle. Traditional linear and nonlinear regression models often fail to predict the popularity of games because of the complex relationship between feature extraction and data. However, because BP neural network and gray prediction have strong fitting ability to nonlinear data, and BP neural network can adapt to different types of data and problems by adjusting network structure and hyperparameters, it is more robust in some cases, and the prediction effect is often more real.