A COMPILER APPROACH TO PHARMACOVIGILANCE

Abstract

Special-purpose compilers had been developed for specific problem domains, which mainly focused on improving certain aspects of the compiler. For instance, the NNVM compiler and the TensorFlow AXL compiler proffers solutions to improve the performance of machine learning algorithms by implementing parallel computing and code optimization to reduce time complexity. An obvious burden in data science/machine learning undertakings is the amount of time needed for data preprocessing. This study has a new look at compiler design using pharmacovigilance as a case study. The research developed a special-purpose compiler to be used in improving data preprocessing as applied to pharmacovigilance. Lexical analysis was applied to data preprocessing and hashing techniques in surveillance and case reporting. The dataset used in the study contains some demographic information of the patients, drugs prescribed, and reported adverse effects. The compiler was built using the Python programming language, and a random forest model was developed using 70% of the data as a training set while the remaining 30% was reserved for testing. The initial model performance in terms of accuracy in reporting adverse events was 0.08; however, after applying hashing techniques and adding the hash as an additional attribute to the dataset, a 1.0 (100%) accuracy was achieved.