Development of Machine Learning Based Recommendation Engine for Embedded Programmer
Yu Zhou, Suxia Cui, Yonghui Wang · 2019
In this paper, we propose a machine learning based recommendation engine that can assist embedded programmers to quickly search and query related code segments and instructions. In the design, we build a code database and each code segment is auto-classified and multi-labeled with the chip model, application scenario, function module, and register name. The similarity of each level is calculated by Hamming Distance, Cosine similarity, and Euclidean distance approaches separately. Based on the different similarities of each level, we come up with a dynamic machine learning model to train the weights of each similarity. The final similarity scores of different codes are generated and sorted, and then the related embedded codes are recommended to programmer according to the final similarity score. Comparisons are performed against single similarity methodology. The experimental results show proposed methodology outperforms the single similarity approach and could assist embedded programmers find the related code more accurately.