Research on Wordle Development Based on Multiple Artificial Intelligence Models

In order to conduct a study of Wordle's development, firstly, forecasts were made using an ARIMA time series analysis model, and the data were further processed using a weighted moving average method so that the forecast values would more accurately reflect future short-term trends. A multiple linear regression model was then chosen and the frequencies of each letter and common letter combination were calculated for the entire dataset. Further, a word cloud was plotted in this paper. By training different models, the linear regression model was found to have the smallest error value and a strong correlation coefficient between letters, letter combinations and number of answers was obtained. Using the strong correlation factors as sample features for multiple linear regression, the trained models were used to accurately predict the word 'eerie'.