Micro-blogging platforms have become the primary channel through which customers publish reactions to food brands, outlets and products, and the resulting stream of short, informal text is a rich but noisy source of consumer opinion. This paper presents TriSent-ML, a five-layer machine learning framework for sentiment classification of food-domain Twitter data. The framework couples an explicit text normalisation layer with a feature construction layer that fuses an n-gram similarity profile with TF–IDF term weighting, and it evaluates three supervised classifiers — Support Vector Machine, Random Forest and Decision Tree — over an identical feature representation so that the contribution of the learner can be isolated from the contribution of the representation. Five algorithms are specified in full, covering normalisation, n-gram profiling, and the three classifiers, together with a complexity analysis of each stage. The framework is evaluated on a publicly available Kaggle corpus of more than 14,000 tweets collected for the KFC and McDonald’s challenge, partitioned into 10,000 training and 4,000 testing instances. The Decision Tree classifier attains the highest accuracy of 88.51%, ahead of Support Vector Machine at 87.71% and Random Forest at 86.55%, while Random Forest returns the strongest precision at 81.67% and the strongest F1-measure at 78.10. Against the existing baseline the framework improves accuracy by between 28.55 and 34.51 percentage points, recall by between 40.80 and 74.67 percentage points, and F1-measure by up to 62.10 percentage points. The results confirm that disciplined normalisation and fused n-gram/TF–IDF weighting dominate classifier selection in this regime, and they establish a strong, computationally inexpensive and fully interpretable baseline against which heavier transformer architectures should be measured.

Kundan Lal Vishwakarma, Prof. Nagendra Patel, "Design and Performance Evaluation of a Multi-Stage Machine Learning Model for Opinion Mining of Food-Related Tweets" International Journal of Scientific Research and Engineering Development, V9(5): Page(306-313) September - October 2026. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.