Early diagnosis of MI is vital to improving patient outcomes in the clinical area. Generally, this work considers using computational intelligence approaches to analyze the clinical texts for the early detection of MI. Introduces MyoCI, which is a novel multimodel integration of linguistic algorithms (LA), principal component analysis (PCA), logistic regression (LR), random forest (RF), and SMOTETomek, to classify transcription texts of MI. Coupled with the use of different LA and AI, a process of MI predictions uses clinical reports to come up with well-informed and enlightening diagnoses. Our approach leverages a varied dataset from mtsamples.com, including transcriptions from multiple medical specialties. The preprocessing pipeline involved text cleaning, lemmatizing, and extracting TF-IDF for feature extraction. The SMOTETomek technique was to deal with the imbalance in classes, and for dimensionality reduction, t-distributed stochastic neighbor embedding was employed. Evaluated various models, including transformer-based ones like BERT, and found out that the RF model with SMOTETomek preprocessing achieved an accuracy of 93.12% and an F1-score of 0.93. Other models, such as LR, and Support Vector Machine, were also assessed.