Breast Cancer Screening Based on Supervised Learning and Multi-Criteria Decision-Making

Abstract

On average, breast cancer kills one woman per minute. However, there are more reasons for optimism than ever before. When diagnosed early, patients with breast cancer have a better chance of survival. This study aims to employ a novel approach that combines artificial intelligence and a multi-criteria decision-making method for a more robust evaluation of machine learning models. The proposed machine learning techniques comprise various supervised learning algorithms, while the multi-criteria decision-making technique implemented includes the Preference Ranking Organization Method for Enrichment Evaluations. The Support Vector Machine, having achieved a net outranking flow of 0.1022, is ranked as the most favorable model for the early detection of breast cancer. The net outranking flow is the balance between the positive and negative outranking flows. This indicates that the higher the net flow, the better the alternative. K-nearest neighbor, logistic regression, and random forest classifier ranked second, third, and fourth, with net flows of 0.0316, −0.0032, and −0.0541, respectively. The least preferred alternative is the naive Bayes classifier with a net flow of −0.0766. The results obtained in this study indicate the use of the proposed method in making a desirable decision when selecting the most appropriate machine learning model. This gives the decision-maker the option of introducing new criteria into the decision-making process.

Evaluating the Performance of Deep Learning Frameworks for Malaria Parasite Detection Using Microscopic Images of Peripheral Blood Smears

“…However, with an accuracy of 96.97%, the proposed model performs significantly better identifying the two classes when microscopic images of thick blood smears are used. Because greater accuracy does not indicate optimum model performance [ 29 ], it is necessary to evaluate the model’s ability to classify infected images as infected and not misclassify uninfected as infected. Further, the harmonic mean of precision and sensitivity is vital as a metric to measure model performance.…”

“…Accuracy is a metric that generally describes how the model performs across all classes. It is useful when all classes are of equal importance [ 29 ]. It is calculated as the ratio of correct predictions to the total number of predictions.…”

Hydrogel-Based Drug Delivery Nanoparticles with Conventional Treatment Approaches for Cancer Tumors; A comparative study using MCDM Technique

“…The evaluation, assessment, and comparison of the most preferred treatment approaches for cancer tumors have always been based on theoretical evaluations and precise experiments. Recently artificial intelligence methods have been deployed for the diagnosis and treatment of cancer and also for comparing diagnostic and therapeutic approaches to identify the most efficient approaches [53]. Many studies in literature have deployed different MCDM techniques to compare, rank, and evaluate different approaches for diagnosing and treating different cancer types.…”

“…The analytic hierarchy process (AHP), a technique for order of preference by similarities to ideal solution (TOPSIS), elimination et Choix traduisant la réalité (ELECTRE), PROMETHEE, visekriterijumska optimizcija I kaompromisno resenje (VIKOR), and data envelopment analysis (DEA) are some of the efficiently applied MCDM methods in different fields [53]. Each has its own set of advantages and disadvantages.…”

Evaluation of the Treatment Alternatives for Spinal Cord Tumors Using Analytical Evaluation Models

“…The idea was to create a middle ground in the presence of uncertainty when making di cult decisions that have possibilities that differ from yes or no [1][4] [38]. Therefore, F-PROMETHEE is a reliable decision-making tool to compare well-structured linguistic fuzzy data [4] [35], [36], [39] [40], [41]. The application of F-PROMETHEE is well detailed in a study by Ozsahin et al [4].…”