Scalable Multimodal Machine Learning for Cervical Cancer Detection

Cervical cancer poses a significant global health concern, and machine learning techniques have shown great promise in early detection and diagnosis. However, the dominant focus on accuracy has overshadowed the crucial aspect of training time. This research paper investigates multimodal high-performance machine learning for cervical cancer detection, with a primary emphasis on scalability and training time optimization using Apache Spark. We conduct experiments on two distinct datasets CSV and Images - analyzing the trade-offs between training time and accuracy. Moreover, we discuss the challenges encountered during the implementation of Spark. Our study helps to shed light on Spark's potential benefits and limitations in high-performance machine learning for cervical cancer detection. Our findings indicate that Spark ML showcases remarkable scalability potential compared to traditional platforms like SkLearn for text datasets. The majority of the ML algorithms explored showcased exceptional accuracy rates, with many surpassing the impressive threshold of 98 % in text data. In particular, we have shown that, Spark helps to reduce the time complexity more than three folds when the data size increases.