International Journal of Trend in Scientific Research and Development (IJTSRD) Volume 5 Issue 6, September-October 2021 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470 @ IJTSRD | Unique Paper ID – IJTSRD46400 | Volume – 5 | Issue – 6 | Sep-Oct 2021 Page 230 Health Risk Prediction Using Support Vector Machine with Gray Wolf Optimization in Covid-19 Pandemic Crisis Swati Shilpi 1 , Dr. Damodar Prasad Tiwari 2 1PG Scholar, 2Assistant Professor, 1,2Department of CSE, BIST, Bhopal, Madhya Pradesh, India ABSTRACT The opinion of disease is important for Covid 19 as the antigen kit and RTPCR are unperfect and should be better for diagnosing such disease. Real-Time Return Transcription (real-time converse transcription – polymerase chain). Healthcare practices include the collection of various sorts of patient data to help the physician diagnose the patient's health. These data could be simple symptoms, first diagnosis by a doctor, or an in-depth laboratory test. These data are therefore used for analyses only by a doctor, who subsequently uses his particular medical skills to found the ailment. In order to classify Covid 19 disease datasets such mild, middle and severe diseases, the proposed model utilizes the notion of controlled machine education and GWO-optimization to regulate if the patient is affecting or not. An efficiency analysis is calculated and compared of disease data for both algorithms. The results of the simulations illustrate the effective nature and complexity of the data set for the grading techniques. Compared to SVM, the suggested model provides 7.8 percent improved prediction accuracy. The prediction accuracy is 8% better than the SVM. This results in an F1 score of 2 percent better than an SVM forecast. KEYWORDS: Covid-19, Pneumonia, Machine Learning, Artificial Intelligence, Healthcare How to cite this paper: Swati Shilpi | Dr. Damodar Prasad Tiwari "Health Risk Prediction Using Support Vector Machine with Gray Wolf Optimization in Covid-19 Pandemic Crisis" Published in International Journal of Trend in Scientific Research and Development (ijtsrd), ISSN: 2456- 6470, Volume-5 | Issue-6, October 2021, pp.230-234, URL: www.ijtsrd.com/papers/ijtsrd46400.pdf Copyright © 2021 by author(s) and International Journal of Trend in Scientific Research and Development Journal. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0) (http://creativecommons.org/licenses/by/4.0) I. INTRODUCTION HRA is one of the critical pieces of custom drug that assesses the general strength of an individual and the shot at having a given sickness. Given the procedure with relationship between fundamental thought subject matter experts and patients, an assistant would be ordinarily prepared to ensure proper HRA gathering. Crucial HRA data covers economics, lifestyle, history and physiological data for individual and family prosperity, (for instance, beat, weight, cholesterol, etc) For example, the USPSTF proposes various procedures for the countering of chest sickness danger, recalling different degrees of information for family parentage. In specific conditions the circuit of genetic information might chip away at additional risk assessments and possibly further foster countering, assumption and treatment. The PALB2 is an amazing portrayal of how to refine the risk information, in this model family parentage, by planning inherited data with standard peril information. Given the significance of fundamental thought in keeping an eye on the overall success of the patient, convincing evaluation of prosperity danger and family parentage will probably further develop risk layering and joint clinical decisions with other clinical providers. This will further develop clinical consideration transport across the prosperity system. II. PREVIOUS WORK There are a couple of papers which have been considered and insinuated on my work. Covid Disease 2019 (COVID-19) spread internationally in mid-2020, making the world face an existential wellbeing emergency. Mechanized recognition of lung contaminations from registered tomography (CT) pictures offers an incredible potential to expand the conventional medical services technique for handling COVID-19. In any case, sectioning contaminated districts from CT cuts faces a few difficulties, remembering high variety for disease attributes, and low force IJTSRD46400 International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD46400 | Volume – 5 | Issue – 6 | Sep-Oct 2021 Page 231 contrast among contaminations and ordinary tissues. Further, gathering a lot of information is illogical inside a brief time frame period, restraining the preparation of a profound model. To address these difficulties, a novel COVID-19 Lung Infection Segmentation Deep Network (Inf-Net) is proposed to naturally recognize contaminated locales from chest CT cuts. In our Inf-Net, an equal incomplete decoder is utilized to total the undeniable level highlights and produce a worldwide guide. Then, at that point, the implied turn around consideration and express edge consideration are used to show the limits and improve the portrayals. Also, to reduce the deficiency of named information, we present a semi-directed division system dependent on an arbitrarily chosen proliferation methodology, which just requires a couple of marked pictures and use fundamentally unlabeled information. Our semi-administered structure can improve the learning capacity and accomplish a better. Broad tests on our COVID SemiSeg and genuine CT volumes exhibit that the proposed Inf-Net beats most state of the art division models and advances the best in class execution (Deng-Ping Fan, Tao Zhou, Ge-PengJi, Yi Zhou, Geng Chen, Huazhu Fu, JianbingShen and Ling Shao; 2020) Covid illness 2019 (COVID-19) is a pandemic brought about by novel Covid. Coronavirus is spreading quickly all through the world. The highest quality level for diagnosing COVID-19 is converse record polymerase chain response (RT- PCR) test. Notwithstanding, the office for RT-PCR test is restricted, which causes early conclusion of the illness troublesome. Effectively accessible modalities like X-beam can be utilized to recognize explicit indications related with COVID-19. Pre- prepared convolutional neural organizations are generally utilized for PC helped identification of illnesses from more modest datasets. This paper examines the viability of multi-CNN, a mix of a few pre-prepared CNNs, for the robotized identification of COVID-19 from X-beam pictures. The strategy utilizes a mix of highlights extricated from multi-CNN with connection based element choice (CFS) method and Bayesnet classifier for the expectation of COVID-19. The strategy was tried utilizing two public datasets and accomplished promising outcomes on both the datasets. In the first dataset comprising of 453 COVID-19 pictures and 497 non-COVID pictures, the technique accomplished an AUC of 0.963 and an exactness of 91.16%. In the second dataset comprising of 71 COVID-19 pictures and 7 non- COVID pictures, the technique accomplished an AUC of 0.911 and an exactness of 97.44%. The trials acted in this investigation demonstrated the viability of pre-prepared multi-CNN over single CNN in the discovery of COVID-19. (Bejoy Abraham, Madhu S. Nair; 2020) This paper proposes a three-stage Susceptible- Infected-Recovered-Dead (3P-SIRD) model to compute an ideal lockdown period for some particular topographical locales that will be ideal to break the transmission chain as well as will assist country's economy with recuperating and backing foundation in a battle against COVID-19. Proposed model is novel since it moreover incorporates boundaries for example quiet transporters, friendliness of recently contaminated individual and unregistered kicked the bucket Covid tainted individuals alongside the disease rate, suspected rate and demise rate. These boundaries contribute a great deal to sort out the clearer model, alongside fundamental boundaries. The model takes the testing pace of suspected individuals into thought and this rate differs as for period of the scourge development. Proposed 3P-SIRD model is partitioned into three-stages dependent on the mindfulness and supportability of illness. Time is separated into various periods as pace of disease and recuperation vacillates district to area. The model is tried on China information and is sufficiently productive to propose a model near their real figures of contaminated individuals, recuperated individuals, kicked the bucket and dynamic cases. The model predicts the ideal lockdown time frame as 73 days for China which is near their genuine lockdown period (77 days). Further, the model is carried out to foresee the ideal lockdown time of India and Italy. (SoniyaLalwani, GunjanSahni, BhawnaMewara, Rajesh Kumar; 2020) In this paper, we research the continuous elements of COVID-19 in India after its rise in Wuhan, China in December 2019. We examine the impact of cross country lockdown executed in India on March 25, 2020 to forestall the spread of COVID- 19. Vulnerable Exposed-Infectious-Recovered (SEIR) model is utilized to gauge dynamic COVID-19 cases in India thinking about the impact of cross country lockdown and conceivable expansion in the dynamic cases after its expulsion on May 3, 2020. Our model predicts that with the continuous lockdown, the pinnacle of dynamic contaminated cases around 43,000 will happen in the mid of May, 2020. We likewise anticipate a 7 to 21% increment in the pinnacle worth of dynamic tainted cases for an assortment of speculative International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD46400 | Volume – 5 | Issue – 6 | Sep-Oct 2021 Page 232 situations mirroring a general unwinding in the control systems carried out by the public authority in the post-lockdown time frame. For India, it is a significant choice to think of a non-drug control procedure, for example, cross country lockdown for 40 days to delay the higher periods of COVID- 19 and to stay away from serious burden on its general medical services framework. As the continuous COVID-19 flare-up stays a worldwide danger, it is a test for every one of the nations to concoct compelling general wellbeing and regulatory techniques to fight against COVID-19 and support their economies. (ChintamaniPai, AnkushBhaskar, VaibhavRawoot) III. PROBLEM IDENTIFICATION The recognized issue in existing work is according to the accompanying:
The chances of identification of Covid 19 patients may lack due to low precision.
Patients recovery is quite low due to obtaining limited F1-Score and Accuracy. IV. METHODOLOGY 1. The proposed prediction model SVM-GWO (Support Vector Machine with Grey Wolf Optimization) method consists of: a) Create a new (N+1)-dimensional input dataset (xT,c)T with N input features [xi,...,xN]T and one output class c. b) You may do this by multiplying the mean of each feature fi by the standard deviation of each feature fi. 2. Implementation of Interactive Computer Aided Design Apply the ICA algorithm to the new dataset, and save the weight matrix W of dimension (N+1) (N+1). 3. Shrinkage of Small Weights A. Calculate the absolute mean for each N+1 independent row vector Wi of W. B. In case |wij| is less than or equal to ai, decrease |wij| to zero. As you can see from the above, is a tiny positive number. 4. Extraction of candidate features A. Create an N-dimensional row weight vector W'i for each weight vector Wi by projecting it over the original input feature space (i.e., deleting weights wi,N+1) that correspond to the output class). B. Create a (N+1)-dimensional vector by multiplying new weight matrix W' of dimension (N+1) N by the original input data x. The components fi's of this vector are new feature possibilities. 5. Removing unsuitable features A. Formulate F = W'i x 1 ••• N+1 as a list of feature candidates. Set FS to F. B. When a feature candidate fi's weight for class wic is 0, then it should be excluded from FS; C. For each feature candidate fi, if corresponding weights wij = 0 for all j ∈ 1 · · · N, then exclude fi from FS. D. It also incorporates final N' extracted features in its FS output. 6. Calculate a decision function using the following parameters as predictors. Fs = Number of vectors Nsv = Number of Support Vectors Nft = Number of features in support vector SV[Nsv] = Support Vector Array IN[Fs] = Input Vector Array F = Decision Function Array for 1 i ← to Fs by 1 do F = 0 for 1 j ← to Nsv by 1 do dist = 0 for 1 k ← to Nft by 1 do dist += (SV[j].feature[k] – IN[i].feature[k])2 end ( ) * exp [j]. dist F SV κ γ α κ = − × + = × end F = F + b* end V. RESULTS AND ANALYSIS The following observations are collect during process of proposed model on patient dataset. Accuracy, Precision and F1-Score parameters are calculate as follows: Table 1: Estimation of Accuracy in between of SVM and Proposed Prediction Model Import Data SVM SVM-GWO (Proposed) 200 0.4 0.51 400 0.53 0.57 600 0.49 0.53 800 0.57 0.61 1000 0.51 0.55 International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD46400 | Volume – 5 | Issue – 6 | Sep-Oct 2021 Page 233 Figure 1: Graphical Analysis of Accuracy in between of SVM and Proposed Prediction Model The above graph show that the proposed model gives better prediction accuracy as compare than SVM. When sample data size is 200 then accuracy improve by 27.5%. In a similar way, when sample data is 1000 then accuracy improve by 7.8%. Table 2: Estimation of Precision in between of SVM and Proposed Prediction Model Import Data SVM SVM-GWO (Proposed) 200 0.41 0.48 400 0.52 0.56 600 0.48 0.52 800 0.56 0.6 1000 0.5 0.54 Figure 2: Graphical Analysis of Precision in between of SVM and Proposed Prediction Model The above graph show that the proposed model gives better prediction precision as compare than SVM. When sample data size is 200 then precision improve by 17%. In a similar way, when sample data is 1000 then accuracy precision by 8%. Table 3: Estimation of F1-Score in between of SVM and Proposed Prediction Model Import Data SVM SVM-GWO (Proposed) 200 0.4 0.46 400 0.57 0.59 600 0.53 0.56 800 0.58 0.61 1000 0.52 0.53 Figure 3: Graphical Analysis of F1-Score in between of SVM and Proposed Prediction Mode. The above graph show that the proposed model gives better prediction F1 score as compare than SVM. When sample data size is 200 then F1 score improve by 15%. In a similar way, when sample data is 1000 then F1 score improve by 2%. VI. CONCLUSIONS The proposed model gives preferred forecast exactness as analyze over SVM. At the point when test information is 1000 then exactness improve by 7.8%. The proposed model gives preferable forecast accuracy as look at over SVM. At the point when test information is 1000 then exactness accuracy by 8%. The proposed model gives better forecast F1 score as think about than SVM. At the point when test information is 1000 then F1 score improve by 2%.Consequently, characterization of patients according to Covid-19 illness indications are better arranged through proposed strategy SVM-GWO (Support Vector Machine with Gray Wolf Optimization). Our proposed philosophy assists with working on the exactness of analysis and enormously accommodating for additional treatment. In future improvements, the exactness must be tried with various dataset and to apply other AI calculations to check the precision assessment. The impediment of the proposed model is handling time, on account of tremendous measure of information taken for assessing the exhibition of train information. In future, similar calculations to be carried out with continuous information for assessing the adequacy of the framework. REFERENCES [1] S. H. A. Khoshnaw, M. Shahzad, M. Ali, and F. Sultan, “A quantitative and qualitative analysis of the COVID–19 pandemic model,” Chaos, Solitons and Fractals, 2020, doi: 10.1016/j.chaos.2020.109932. [2] WHO, “Novel Coronavirus (2019-nCoV),” WHO Bull., 2020. International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470 @ IJTSRD | Unique Paper ID – IJTSRD46400 | Volume – 5 | Issue – 6 | Sep-Oct 2021 Page 234 [3] A. Waris, U. K. Atta, M. Ali, A. Asmat, and A. Baset, “COVID-19 outbreak: current scenario of Pakistan,” New Microbes and New Infections. 2020, doi: 10.1016/j.nmni.2020.100681. [4] N. Noreen et al., “Coronavirus disease (COVID-19) Pandemic and Pakistan; Limitations and Gaps,” Limitations Gaps. Glob. Biosecurity, 2020. [5] J. Wu et al., “Rapid and accurate identification of COVID-19 infection through machine learning based on clinical available blood test results,” 2020, doi: 10.1101/2020.04.02.20051136. [6] H. Yue et al., “Machine learning-based CT radiomics method for predicting hospital stay in patients with pneumonia associated with SARS-CoV-2 infection: a multicenter study,” Ann. Transl. Med., 2020, doi: 10.21037/atm- 20-3026. [7] R. Kumar et al., “Accurate Prediction of COVID-19 using Chest X-Ray Images through Deep Feature Learning model with SMOTE and Machine Learning Classifiers,” pp. 1–10, 2020, doi: 10.1101/2020.04.13.20063461. [8] R. M. Pereira, D. Bertolini, L. O. Teixeira, C. N. Silla, and Y. M. G. Costa, “COVID-19 identification in chest X-ray images on flat and hierarchical classification scenarios,” Comput. Methods Programs Biomed., 2020, doi: 10.1016/j.cmpb.2020.105532. [9] X. Xu et al., “Deep learning system to screen coronavirus disease 2019 pneumonia,” arXiv. 2020. [10] D. Brinati, A. Campagner, D. Ferrari, M. Locatelli, G. Banfi, and F. Cabitza, “Detection of COVID-19 Infection from Routine Blood Exams with Machine Learning: A Feasibility Study,” J. Med. Syst., 2020, doi: 10.1007/s10916-020-01597-4.