Diagnosis of cervical cancer using effective hybrid model in Bangladesh

Abstract

This paper discusses the various forms of restraints, regulations, struggles, and reme dies for overcoming cervical cancer challenges. It is a life-threatening disease that affects women all over the world. The lack of adequate early treatment has been one of the leading causes of this. As a result, early diagnosis of cervical cancer is important. A substantial number of deaths in Bangladesh were caused by a short age of specialists for cervical screenings. In addition, hospitals in many parts of Bangladesh do not have medical equipment for cervical screening, such as Pap smear, colonoscopy, and biopsy. An au-dominated system can therefore be extremely valu able for quick identification and treatment if the likelihood of cervical abnormalities is predictable. We employed hybrid machine education for this job. From Hybrid Machine Learning, we worked with K-Nearest Neighbor, Decision Tree, Naive Bayes, Random Forest, Support Vector Machine, Stochastic Gradient Descent, Ada- boost, and Recurrent Neural Network to find the data classification. KNN uses architec tures for easy and accurate detection including the classification of cervical cells. With the use of KNN, we use demographic data to determine whether or not a pa tient has an atypical cervix instead of picture data. The advantage of a decision tree is that all potential decisions will be analyzed and each path concluded. Moreover, we have used the Naive Bayes algorithm. It is a probabilistic machine that may be used for a variety of classification tasks. Random forest is a supervised learning algorithm designed to solve classification problems. It’s just a set of decision trees whose outcomes are combined into a single final result. We have also used SVM in the prediction of cervical cancer. RNN is a type of artificial neural network in which nodes’ connections form a directed graph that follows a temporal series. It can now show temporal hierarchical actions as a result of this. We have applied for RNN and got a good result. In each iteration, Ada-boosts are used to set the classifier weights and to train the data sample so that those odd observations are predicted accurately. Thanks to a unique training example processed by the network, SGD is easy to match to memory. It is quick as only one sample is analyzed simultaneously. These models and methods have achieved higher accuracy of the decisions that they recommend including deep understanding to make decision-makers easier to imple ment. We got better accuracy by learning implicit, non-implicit, and non-symbolic knowledge. Furthermore, in the future, we will use more algorithms to improve accuracy.