2.1. Based on Risk Factors for Cervical Cancer

The “National Comprehensive Cancer Network” has issued a warning about the benefits of initial identification of cervical cancer. In contrast, a postponement in treatment is the leading cause of an increasing number of women mortality globally. As a result, numerous scientific and medical investigations have investigated the causes, symptoms, and methodologies of identifying and avoiding cervical cancer. Researchers have also attempted to evaluate the risks that contribute to the pathogenesis and progression of this particular cancer. The selected research works are as follows.

In the research article [7], the cure for cancer has usually taken numerous forms over the years; total elimination may not even be possible; however, the disease’s probability of occurrence and forecasting can be reduced. Any disorder can be healed if identified in its beginning phases, and cancer can be successfully treated if spotted in its beginning phases. On the other hand, cervical cancer is hard to forecast in its early stages because there are no symptomatic. The frequent test is done for such forecasting of cancer cells because testing has been the only way it can be forecasted [8]. In [9], to avoid such uncertainties, screening outcomes may be supervised as false positives at points in time, or they may be postponed. Machine learning has been developed in the field of health care services. Numerous methods, techniques, and technology have been used to anticipate cancer cells quicker and with a lower false-positive rate.

The method of mathematical modeling aids in the comprehension of the observable occurrence. The visible event in the healthcare area [10] could be wellness symptoms and perhaps a sickness, and this technique results in a workable characterization of complicated things. Inside the medical sciences, the mathematical formulation has also been utilized in various methods to solve, reproduce, research, and explain biological mechanisms [11]. The research [12] proposes probabilistically mathematical systems when the sample sizes are limited and can thoroughly examine the parameters. According to the researchers, any healthcare system may comprehend via comparisons; then, such a procedure must influence the mathematical framework [13]. As illustrated, a model named three separate structures might be used to understand the number of carbohydrates stored in human bodies. Other researchers prefer to use informative computational methods. These models use a feasible description of factors in analytics testing to describe realistic circumstances [14]. In social and epidemiology investigations, description methods are essential. In most cases, the means, median, average, standard deviation and variance, and other statistics are determined, and a report of the phenomena is written down. Table 1 represents the summary of existing research work based on cancer risk factors.

2.2. Based on Mathematical Models

Furthermore, more examination into cervical cancer using mathematical models indicates that significant teams of investigators in the medical sciences concentrate on diagnostics modeling models [28]. The experts in clinical forecasting use a variety of strategies to construct models. Analysis technique and supervised learning model are two examples. Specific healthcare computer models are referred to as “forms of modern.” Basic logical reasoning, hypotheses, concepts, and descriptive analysis have created these frameworks. Many researchers usually refer to such algorithms as medical condition recognition systems [29]. They also utilized ML algorithms to predict serious health issues by the researchers. Enzyme kinetics and pharmacokinetics are two necessary fields of medical research [30]. Machine learning algorithms and automatic analyses are frequently used in several areas of medicine. Physiological reactions and parameters like stress levels, heartbeat, and others must be recorded and modeled for tracking medical conditions within time-series modeling techniques [31]. Modeling, which enables to comprehension of dynamic interaction, uses an approach called transferring characteristics for a detailed look. This type of procedure keeps track of feedback and the processes between this. Many researchers have looked at the principal source of such medical conditions while discovering and establishing the mathematical determinant factors.

Nevertheless, the issue is mainly identifying acceptable factors that can describe the specialized clinical paradigm or phenomenon and determining which independent variables may operate as potential forecasters and which characters can describe the entire computational formula [32]. All of the clinical models presented thus far depend on a fundamental grasp of the mathematical model development. Depending on the concerns and obstacles described in the present research, this next section considers the frame of the activity. Table 2 represents the review of cancer types based on several features and age group impact.

2.3. Based on Machine Learning Models

In this research, machine learning techniques have been employed to detect cervical cancer accurately via constructing a framework affected by previous research methods in a similar domain. Research [42] proves that by utilizing the oversampling process performance of existing approaches can be improved. This research used the random forest to build a classifier predicated on cervical cancer cases. The analysis indicates that the RF significantly outperformed its same framework after implementing SMOTE, including all characteristics of cervical disease variables in the forms of parameters, i.e., accuracy, specificity, precision, and true positive rate. The research [43] used the online UCI dataset with various strategies for cervical cancer diagnosis: (a) SVM, (b) SVM with PCA, and (3) SVM with RFE. This article concluded that SVM performs well and achieves better precision, diagnostic accuracy, and precision than the multiple different classifiers.

Research [44] utilized three forms of machine learning models to categorize the UCI cervical cancer data. The proposed model used a “border row hierarchical clustering” (BRHC) to deal with dataset inequity. This research has observed that the XG-Boost and random forest methods perform outstandingly in cancer prediction accuracy rates. Since this cancer data contains many incomplete, missing data, it is necessary to deal with missing attributes carefully. Research [45] offers four distinct methods to deal with missing values in the cancer dataset. These techniques are NOCB, LOCF, FVM, and NOCB. To anticipate the biopsy input variables, they utilized six algorithms: LR, RF, SVM, DT, NB, and NN [46], and researchers also concluded that if used with the NOCB preprocessing phase, the SVM, as well as LR, reached the best accuracy, F1 measure, and precision. In this research, machine learning techniques have been employed to detect cervical cancer accurately via constructing a framework affected by previous research methods in a similar domain. The private database was created using 472 survey questions from a China health center, so each cancer patient who took the poll had a correlating gene sequence set of data. This research collects the data from “Mexico’s Maggiore de Caracas health center.” This dataset contains 592 cancer patients’ data with various attributes. This research applied a pooling and discussed the difficulties associated with conventional cervical cancer diagnostics. Table 3 represents the comparison of research methods based on ML methods.

Machine learning approaches have been utilized in this investigation to correctly identify cervical cancer via developing a structure influenced by prior research methodologies used in a similar field. The public available UCI dataset on cervical cancer does not have per-annotated rows that give a confirmatory signal about the presence or absence of cervical cancer. The dataset aims to understand the subjects that influence a cervical cancer diagnosis.