Prognosis models for nasopharyngeal carcinoma recurrences by using tabu search algorithm

Cancer is a significant public health issue that has a global impact. Significant mortality rates have already been observed due to this disease, and more mortalities are expected in the future. In recent times, there has been a growing interest among otolaryngologists and oncologists in the development of appropriate treatment regimens for patients with recurrent nasopharyngeal carcinoma (NPC). The primary objective of these treatment modalities is to extend the lifespan of patients following recurrence and enhance their overall survival and quality of life. For instance, metaheuristic algorithms (MH), a form of soft computing technology, are commonly utilized in healthcare data due to their effectiveness. Furthermore, metaheuristics rely on the evolutionary search principle. They direct the search process to effectively explore the search space in order to find near-optimal solutions for solving global optimization problems. Tabu search (TS) is a method used in optimization problems and falls under metaheuristic techniques. An essential element of TS is its utilization of adaptive memory, which enhances search efficiency by avoiding local optimality and promoting flexibility. Another example is data mining, which is a subset of artificial intelligence that utilizes data to extract meaningful information from previously unknown patterns. It has been increasingly used in healthcare to aid clinical diagnostics and disease prediction. The proposed technique treated data mining problems as combinatorial optimization problems and used metaheuristics to address data mining challenges, such as classification for unknown data and finding association rules for significant patterns. The Tabu Search Classifier Method (TSCM) outlined in this paper primarily utilizes the Tabu Search (TS) algorithm, enhanced with the incorporation of Dynamic Neighborhood Structure (DNHS), which contributes to better discovery of the search space. The TSCM algorithm identifies three rules based on the patients’ data and generates three precise artificial predictive models to determine and categorize individuals who are at risk of recurrent NPC. With each stage of the treatment, additional features become accessible. The first model relies on a primary data set that includes descriptive data. The second model incorporates more features than the first model but does not include the response feature. The third model utilizes all existing features and includes the response feature, which is observed three months after the treatment phase concludes, the third model is considered a post-treatment monitoring. This paper introduces an Artificial Advisory Healthcare System (AAHS) that utilizes these models to accurately predict the occurrence of recurrence during each stage of treatment and after the treatment as a post-treatment monitoring. This prediction enables the adjustment of the treatment plan and the implementation of additional measures in accordance with the system’s outputs. Given the growing prevalence of artificial intelligence in medical research, the proposed system aims to predict the likelihood of NPC recurrence in certain patients. Therefore, this will enable oncologists to take additional medical precautions to prevent recurrence and enhance their understanding of cancer. The experimental results indicate that the three proposed predictive models outperform the existing prognoses for NPC recurrence.