Author (s) :

---

1). F.O Aranuwa, Adekunle Ajasin University, Akungba Akoko, Ondo State, NIGERIA
2). O. J. Popoola, Adekunle Ajasin University, Akungba – Akoko, Ondo State, NIGERIA

Abstract :

---

Long Short-Term Memory (LSTM) network, unlike traditional feed forward neural networks is capable of processing complex and sequential data with minutest error. This classifier possessed feedback connections and its activation patterns in the network change once per time-step, and this characteristics make its algorithm ideal for classification and prediction problems. This study designed a model for efficient classification and prediction of prostate cancer and prostate enlargement using the LSTM network. Prostate cancer and prostate enlargement, also known as benign prostatic hyperplasia (BPH) relatively shared similar risk factors and symptoms that makes their classification and prediction a complex task. Researchers have made efforts to address these issues, however the sequence prediction problems still draw more research attentions. The inability of the classical methodologies to learn and synthesize underlying relationships among the symptoms, always results in low predictive accuracy and high error rates. Data attributes for the study were sourced from selected medical institutions in the South Western, Nigeria. A total of 1,149 datasets were collected, prepared and made suitable for modeling through transformation and normalization processes. Samples were categorized based on specific attribute determinants, whose results could be Cancer, BPH and Normal (1, 2, 0). The dataset was trained and tested on 80/20 ratio on Anaconda platform. The experiment was set to run on 70 epochs, to determine the accuracy and ascertain the model’s efficiency and practicality. Results shown that the model correctly classified and predicted the status of the malaise with 96.7% accuracy over the benchmark of 92%. The model, haven suitably addressed the sequence prediction problems, which is an improvement on the previous models considered, is recommended for deployment in various health care sector for efficient classification and prediction of the disease.

---