Nowadays, many people, especially youth, express their opinions regarding different topics using tweets. It has become essential to extract information from the tweets posted by users. This information is helpful for decision-making to increase business profit and to identify and stop the spreading of wrong or sensitive information in critical situations. It is observed that the majority of tweets express multiple emotions, thereby making use of multi-label (ML) classification crucial for processing data in them. MLCET, which takes preprocessed tweets as input and then uses prior and estimated probabilities computed from statistics of neighbors using feature similarities and label dissimilarities, performs better than three ML algorithms when evaluated for five metrics. The work also depicted the importance of stemming and lemmatization for enhancing performance on text data. Experimentation has shown 9%, 16%, 15%, 23%, and 16% improvements in MLCET for one error, accuracy, F1 score, macro, and micro F1, respectively.