Wordle difficulty prediction model based on PSO-LSTM and hierarchical clusters

This paper focuses on word difficulty prediction and classification on the basis of capturing a year's word game data. Although Wordle was all the rage, recent tracking data shows that the Wordle popularity appears to be declining gradually, and the number of players has a downward trend. To promote the healthy development of Wordle and assist game developers in adjusting their game strategies in time, this paper establishes a word difficulty prediction model. Before Wordle difficulty prediction model is established, we preprocessed the data, rasterizing the area to be studied. In phase I, we introduced a PSO-LSTM prediction model and performed specific training, achieving an accuracy rate of over 85%. After coding the letters in "EERIE", we successfully obtained the percentage of word "EERIE" on March 1,2023 will appear. The second phase uses hierarchical clustering analysis to subcategorize words by difficulty. The difficulty of words is then divided into five levels based on the folded graph of clustering coefficients derived from the elbow rule. Finally, this paper presents a model accuracy test and suggests the idea that this model can predict the data of serial relationship within the existence time and can be extended to the evaluation and weighted ranking under the influence of other factors.