Predictive Analysis of Wordle Games Based on BP Neural Networks and K-Means Clusters

Wordle is a popular daily puzzle game currently offered by The New York Times in of which players can guess words and receive feedback on each guess. Wordle games are currently growing in popularity, but research into Wordle games is still very limited and there are many issues that need to be examined in depth.First,We need to build models for different word attributes and predict the distribution of the results reported for any one word, giving the predicted results for the word “EERIE” on March 1, 2023. Here we first processed the data to get the final outcome data, then we introduced a BP neural network prediction, we introduced a BP Neural Network prediction model and performed specific training, where the training accuracy reached over 98%, which is very accurate. We then filtered the features using XGBoost and we created a difficulty classification based on a K-means clustering analysis that divided the difficulty into two categories: difficult and easy, we then performed specific clustering and obtained the final clustering results for evaluation. Secondly, we created a linear discriminant analysis of classification accuracy and performed statistics for each word attribute, which turned out to be very consistent with our previous analysis. The more common the word is in life, the lower the difficulty. The more the word repeated letters, the higher the difficulty. Finally, we introduced the EERIE words into the clustering model and found them to be of medium difficulty and got better results.