Peptides that are efficient against bacteria, fungi, and viruses are known as antimicrobial peptides. Feature vectors generated by the vast majority of machine learning approaches are always the same length, even though the number of amino acids in various peptides could vary widely. It is common practise to select features that are not optimal for the task at hand since there is a lack of direction regarding whether or not the features reflect periodic patterns in the peptide sequence that are significant to the classification issue that is now being faced. As a consequence of this, the product is permitted to contain a sizeable amount of filler we build a feature vector by constructing feature representations of individual amino acids. This allows us to tackle the challenges we were facing. This indicates that there will be no surprises regarding the size of the final feature vector. When it comes to the classification of antimicrobial peptides, this study evaluates the performance of k-nearest neighbour classifiers, Random Forest classifiers, and LSTM Recurrent neural networks to see which yields the most accurate results.