Analyzing and Predicting Player Behavior by SARIMAX and Neural Networks

This article focuses on the trend analysis of Wordle puzzles and predicting player behavior through advanced models. This paper found based on the number of wordle users on social platforms in 2022 that this number grows in the early burst and then the region declines smoothly. This paper introduce an exogenous variable-word difficulty-to optimize the model, The model that has been changed is called SARIMAX. The predicted player count for March 1, 2023 is between 15,647 and 29,059. Our forecast distribution for the word ‘EERIE’ on March 1, 2023 is 2 to 4 occurrences, with a likelihood of 89%. This article choose to fuse the word difficulty evaluation, calculated from dataset, with the classical K-means clustering algorithm to classify the words into 5 classes according to their difficulty. The model introduces a neural network classification algorithm that eventually classify words based on their own properties with high accuracy. Substituting EERIE into this model, the word was found to be of medium difficulty.