View of Enhanced Deep Learning Assisted Convolutional Neural Network for Heart Disease Prediction

(1)

8467 http://annalsofrscb.ro

Enhanced Deep Learning Assisted Convolutional Neural Network for Heart Disease Prediction

Dr.V. Vennila

¹

, T. Akillesh

²

, S. Bharathi

³

, C. Karthik

⁴

, Dr.V.Sharmila

⁵

1Assistant Professor, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamilnadu, India.

2,3,4B.E Student, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamilnadu, India.

5Professor, Department of Computer Science and Engineering, K.S.R College of Engineering, Tiruchengode, Tamilnadu, India.

ABSTRACT

The finding of coronary illness has become a troublesome clinical undertaking in the current clinical exploration. This determination relies upon the definite and exact investigation of the patient's clinical test information on a person's wellbeing history. The gigantic advancements in the field of profound learning try to make canny computerized frameworks that help specialists both to foresee and to decide the sickness .Therefore, the Enhanced Deep learning helped Convolutional Neural Network (EDCNN) has been proposed to help and improve quiet prognostics of coronary illness. The EDCNN model is centered around a more profound design which covers multi-layer perceptron's model with regularization learning draws near. In this way, incited for elective techniques, for example, AI calculations that could utilize non-intrusive clinical information for the coronary illness determination and evaluating its seriousness. Moreover, the framework execution is approved with full highlights and limited highlights.

Subsequently, the decrease in the highlights influences the productivity of classifiers regarding handling time, and exactness has been numerically investigated with test results. The EDCNN framework has been actualized Platform for choice emotionally supportive networks which encourages specialists to viably analyze heart patient's data in cloud stages anyplace on the planet. The test outcomes show contrasted with regular methodologies approaches, for example, Multi-Layer Perceptron's (MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) in view of the examination the planned analytic framework can productively decide the danger level of coronary illness adequately. Test outcomes show that an adaptable plan and ensuing tuning of EDCNN hyper boundaries can accomplish an accuracy.

KEYWORDS

Convolutional Neural Network, EDCNN, MLP.

Introduction

Heart Disease

Coronary illness is an overall term that implies that the heart isn't working typically. Infants can be brought into the world with coronary illness. This is called innate coronary illness. On the off chance that individuals get coronary illness later, it is called procured coronary illness. Most coronary illness is gained.

The three most regular sorts of procured coronary illness are:

Coronary Artery Disease (Abbreviation CAD)

This is an issue with the veins that convey blood to the heart muscle. On the off chance that these veins get minuscule, or in the event that they become obstructed, blood can't move through them typically. Since less blood is provided to the heart muscle, the muscle can't work at typical limit. The heart muscle can get debilitated and frail.

Heart muscle can even bite the dust if blood stream stops. Obstructed corridors in the heart are frequently brought about by smoking, elevated cholesterol, hypertension, diabetes, and acquired qualities from guardians. These issues harm the covering of the heart's veins and cause them to become limited or hindered totally.

Congestive Heart Failure (CHF)

This is a condition that implies that the heart isn't siphoning at typical levels. Two normal causes are a frail or wiped out heart muscle and unusual heart valves. The valves may not let sufficient blood through in light of the fact that

(2)

they are excessively limited. Or on the other hand the valve may "hole" and allow blood to stream in reverse (a misguided course) inside the heart. At the point when the heart valves don't work ordinarily, the heart muscle needs to accomplish additional work and it can get drained.

Bad Heart Rhythms

This is an issue with electrical movement in the heart. This can make the heart beat excessively quick or excessively sluggish. Bad heart rhythms may make the heart quit siphoning blood. The heart needs an ordinary mood to siphon the blood well. In the event that the cadence is too quick, the heart might not possess energy for blood to enter the chambers, so there isn't sufficient blood traveling through the heart with each beat. In the event that the heart is excessively lethargic, there may not be sufficient compressions of the heart to supply the body with the blood that it needs.

Convolutional Neural Network

A convolutional neural organization (CNN, or Conv Net) is a class of profound neural organizations, most ordinarily applied to breaking down visual symbolism. They are otherwise called move invariant or space invariant counterfeit neural organizations (SIANN), in light of their shared-loads design and interpretation invariance attributes. They have applications in picture and video acknowledgment, recommender frameworks, picture grouping, Image division, clinical picture investigation, characteristic language preparing, mind PC interfaces, and monetary time arrangement. CNNs are regularized adaptations of multilayer perceptron. Multilayer perceptron typically mean completely associated networks, that is, every neuron in one layer is associated with all neurons in the following layer. The "completely connectedness" of these organizations makes them inclined to overfitting information.

Commonplace methods of regularization incorporate adding some type of size estimation of loads to the misfortune work. CNNs adopt an alternate strategy towards regularization: they exploit the various levelled design in information and collect more perplexing examples utilizing more modest and less complex examples. Thusly, on the size of connectedness and intricacy, CNNs are on the lower extraordinary. Convolutional networks were propelled by natural cycles in that the availability design between neurons looks like the association of the creature visual cortex.

Individual cortical neurons react to boosts just in a confined area of the visual field known as the open field. The responsive fields of various neurons halfway cover with the end goal that they cover the whole visual field. CNNs utilize moderately minimal pre-preparing contrasted with other picture grouping calculations. This implies that the organization learns the channels that in conventional calculations were hand-designed. This autonomy from earlier information and human exertion in component configuration is a significant preferred position.

Deep Learning

Profound learning (otherwise called profound organized learning) is essential for a more extensive group of AI strategies dependent on fake neural organizations with portrayal learning. Learning can be managed, semi-directed or solo. Profound learning structures, for example, profound neural organizations, profound conviction organizations, intermittent neural organizations and convolutional neural organizations have been applied to fields including PC vision, machine vision, discourse acknowledgment, characteristic language preparing, sound acknowledgment, informal community sifting, machine interpretation, bioinformatics, drug plan, clinical picture examination, material investigation and pre-packaged game projects, where they have created results practically identical to and sometimes marvellous human master execution. Counterfeit neural organizations (ANNs) were enlivened by data preparing and circulated correspondence hubs in natural frameworks. ANNs have different contrasts from organic cerebrums. In particular, neural organizations will in general be static and emblematic, while the natural mind of most living life forms is dynamic (plastic) and simple. The descriptor "profound" in profound learning alludes to the utilization of various layers in the organization. Early work showed that a straight perceptron can't be a general classifier, and afterward that an organization with a nonpolynomial actuation work with one concealed layer of unbounded width can then again so be. Profound learning is an advanced variety which is worried about an unbounded number of layers of limited size, which licenses down to earth application and improved usage, while holding hypothetical all inclusiveness under mellow conditions. In profound learning the layers are likewise allowed to be heterogeneous and

(3)

to veer off broadly from organically educated connectionist models, for proficiency, teachability and understandability, whence the "organized" part.

Related Work

Aleksei Dudchenko, Matthias Ganzinge It could be found in the earlier many years that Machine Learning (ML) has a gigantic assortment of potential usage in medication and can be of incredible use. By and by, cardiovascular sicknesses cause about 33% of the all out worldwide passings. Accomplishes ML work in the cardiology area and what is the current advancement in such manner To respond to this inquiry, we present an orderly audit targeting distinguishing examines where AI calculations were applied in the space of cardiology Providing an outline dependent on the current writing about the best in class ML calculations applied in cardiology. We were unable to give a quantitative gauge for the calculations because of the heterogeneity of the measurements utilized in various examinations A promising part of Machine Learning, the 'Support Learning', was likewise never proposed in the noticed papers. Tree-based groups are normal and show great outcomes, though profound neural organizations are inadequately addressed. Most papers (20 of 27) have utilized datasets that are not really accessible for different specialists, for example unpublished neighborhood vaults. We likewise recognized 28 distinct measurements for model evaluation.[1]

Amin UlHaq Jian Ping It is a major test for the examination local area to build up an analysis framework to distinguish diabetes in a fruitful manner in the e-medical care climate. AI procedures have an arising part in medical services benefits by conveying a framework to examine the clinical information for conclusion of sicknesses Classifier Decision Tree has been utilized for the arrangement of solid and diabetic subjects. The test results show that the proposed highlight choice calculation chose highlights improve the order execution of the prescient model and accomplished ideal precision. Also, the proposed framework execution is high contrasted with the past best in class techniques. Elite of the proposed technique is because of the various blends of chosen highlights set and Plasma glucose focuses, Diabetes family capacity, and Blood mass list are all the more fundamentally significant highlights in the dataset for expectation of diabetes The current finding frameworks have a few disadvantages, for example, high calculation time, and low forecast exactness. 'To deal with these issues, we have proposed a conclusion framework utilizing AI strategies for the identification of diabetes. The proposed technique has been tried on the diabetes informational collection which is a clinical dataset planned from patient's clinical history. Further, model approval strategies, for example, wait, K-overlay, leave one subject out and execution assessment metrics.[2]

U. Haq, J. Li, M. H. Memon, M. H. Memon, J. Khan, and S. M. Marium Detection of Heart Disease (HD) by utilizing models of AI (ML) is exceptionally powerful in beginning phases. The HD treatment and recuperation is compelling whenever recognized the sickness at introductory stages. HD recognizable proof by AI (ML) strategies has been created to help the doctors. In this investigation we proposed an Identification framework by utilizing ML models to characterize the HD and solid subjects. Successive in reverse choice of highlight calculation was utilized to choose more fitting highlights to expand the arrangement precision and decreased the computational season of prescient framework. Cleveland coronary illness dataset was for assessment of the framework. The dataset 70%

utilized for preparing and staying for approval. The proposed framework exhibitions have been estimated by utilizing assessment metrics. In request to conclusion coronary illness a productive analysis framework has been proposed in this examination. SBS include choice calculation was utilized to choose more fitting highlights to expand the characterization exactness and decreased the computational season of prescient framework. Cleveland coronary illness dataset was in this examination and 70% for preparing and 30 % for testing of the dataset.[3]

Senthilkumar Mohan, Chandrasegar Thirumalai, Heart illness is perhaps the main sources of mortality on the planet today. Expectation of cardiovascular infection is a basic test nearby clinical information examination. AI (ML) has been demonstrated to be viable in helping with settling on choices and expectations from the huge amount of information created by the medical care industry. We have additionally seen ML strategies being utilized in late improvements in various zones of the Internet of Things (IoT). Different investigations give just a brief look into anticipating coronary illness with ML strategies. In this paper, we propose a novel strategy that targets finding critical highlights by applying AI procedures bringing about improving the exactness in the forecast of cardiovascular infection. The expectation model is presented with various mixes of highlights and a few known

(4)

characterization procedures. We produce an upgraded execution level with a precision level of 88.7% through the forecast model for coronary illness with the crossover irregular timberland with a straight model (HRFLM).[4]

P. Ramprakash, R. Sarumathi, R. Mowriya Healthcare involves a key part in living souls. The medical care industry contain huge measure of mental information thus AI models were utilized to give end successfully in the coronary illness forecast. The characterization of sound individual and non-solid individual should be possible dependably by utilizing AI techniques. We built up a structure in this investigation that can comprehend the standards of anticipating the danger profile of patients with the clinical information boundaries. The proposed model is developed utilizing Deep Neural Network and 2 - factual model. The issue of under fitting and over fitting is killed. This model shows better outcomes on both the testing and preparing information. DNN and ANN were utilized to examine the proficiency of the model which precisely predicts the presence or nonappearance of heart disease.[5]

V. Vennila, A. Rajiv Kannan has introduced parallel linguistic fuzzy rule with canopy MapReduce (LFR-CM) framework. In this framework canopy MapReduce function is used to classify big data information sharing in the cloud with higher classification accuracy and lesser time consumption.[15]

P. Balamurugan, M. Shyamala Devi, V. Sharmila has introduced the optimized methods for securing data (OMSD) which is for secure data transmission with trust based weights.[16]

P. Balamurugan, M. Shyamala Devi, V. Sharmila has proposed Score-based data gathering algorithm which provides a significant solution to maximize the network lifetime as well as minimum delay per round of data gathering.[17]

V.Vennila, A. Rajiv Kannan proposed Discritized Support Vector Classification and Prediction (DSV-CP) model for sharing information in the cloud environment by using efficient Big Data computation.[18]

V. Sharmila, G. Tholkappia Arasu, P. Balamurugan has proposed a non-class element based iterative clustering approach. Weight calculation are used for selection classes. [19]

P. Balamurugan, T.Ravichandran, V.Sharmila proposed Grade- Based Data Gathering (GBDG) algorithm for minimizing wireless sensor networks energy consumption.[20]

V. Sharmila, P.Balamurugan, V.Vennila, S.Savitha has proposed a data verification scheme. In which malicious data packets are identified.[21]

V. Vennila, A. Rajiv Kannan has proposed Parallel Symmetric Matrix-based Predictive Bayes Classifier (PSM-PBC) model is developed for efficient Big Data computation and information sharing in Cloud environment.[22]

Proposed Methodology

Thus, the decrease in the highlights influences the productivity of classifiers regarding handling time, and precision has been numerically broke down with test results. The EDCNN framework has been executed Platform for choice emotionally supportive networks which encourages specialists to viably analyze heart patient's data in cloud stages anyplace on the planet. The test outcomes show contrasted with ordinary methodologies approaches, for example, Multi-Layer Perceptron's(MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) in view of the investigation the planned indicative framework can proficiently decide the danger level of coronary illness viably. Test outcomes show that an adaptable plan and ensuing tuning of EDCNN hyper boundaries can accomplish an exactness The resulting execution of the profound learning strategies is surveyed for the conclusion of cardiovascular infection regarding execution measures, remembering the likelihood of mistake for the order, indicative precision, accuracy, affectability, particularity.

(5)

Data Visualization and Pre-processing

The Wisconsin Prognostic Cleave Land Train Dataset is downloaded from the UCI Machine Learning Repository site and saved as a content record. This document is then brought into Excel accounting page and the qualities are saved with the comparing ascribes as section headers. The missing qualities are supplanted with fitting qualities. The ID of the patient cases doesn't add to the classifier execution. Henceforth it is eliminated and the result trait characterizes the objective or ward variable subsequently diminishing the list of capabilities size to 33 ascribes. The algorithmic procedures applied for highlight significance investigation and grouping are intricately introduced in the accompanying sections. The dataset has been preprocessed for productive use by the classifier strategies, for example, erase of missing qualities, ordinary scalar, or Min and Max Scalar.

Risk Prediction

Default risk is the chance that associations or individuals will be not ready to make the fundamental portions on their commitment responsibilities. Toward the day's end, credit default risk is the probability that if you advance money, conceivably they will not have the choice to give the money back on time. It gives reproducible and target finding, and henceforth can be a significant aide apparatus in clinical practices. Results are similarly, encouraging and in this manner the proposed technique will be useful in infection diagnostics. To lighten the impact of default peril, banks habitually power charges that identify with the record holder's level of default danger. A more critical degree of threat prompts a higher required return. At that point the information are grouped utilizing Multi-Layer Perceptron's(MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) utilizing all the highlights of CAD information. Danger estimate instruments are made to perceive patients at serious risk and to support specialist dynamic. The result of the assumption models can be used to pick the most fitting/recommended procedure. An examination of displayed that the use of a peril gauge gadget energized specialists in getting more aware of the outcomes, in getting more taught on peril factors and to have a more rousing attitude toward preemptive organization.

Classification Algorithms

Planned profound learning calculations are centered around a profound multi-layer understanding of framework and plan guideline. Further, the analysis design is utilized to identify if patients have coronary illness dependent on the preparation model The presentation has been approved for exactness, the mistake likelihood, particularity, affectability, accuracy. To request sex (target class) using hair length as feature limit, we could set up a model using any gathering counts to compose some plan of breaking point conditions that can be used to isolate the male and female genders using hair length as the planning incorporate. In sex portrayal case the cutoff condition could the most ideal hair length regard. At that point the information are bunched utilizing Multi-Layer Perceptron's(MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) utilizing all the highlights of CAD data.

Further, a distant patient checking (RPM) stage is proposed, that is sufficiently handy to screen the patient commonly with help To gather data about the patients' wellbeing boundaries, for example, heartbeat, ECG and circulatory strain and send an emergency cautioning to the guardian with their genuine condition and complete healing subtleties. Here the Feature determination is required for profound learning help in light of the fact that occasionally non-important highlights influence the profound learning order effectiveness. The choice of highlights builds the exactness of order and decreases the model time. The DL calculations have been utilized for choosing highlights, and a multi-layer perceptron calculation has been used for twofold order issues.

Ensemble Classification Report

EDCNN has been proposed for the early forecast of coronary illness and analysis. The UCI archive dataset has been used for the finding reason, and CNN classifier and multi-layer perceptron (MLP) module has been utilized to order essential ECG pulses for include extraction. The CNN capacities as an element extractor block because of the beat characterization issue. The last initiations getting from the last convolution layer are utilized as contributions to an organization. A cluster standardization layer and an enactment work follow the fundamental convolutional layer utilizing a numerical convolutional measure We use the arrangement dataset as far as possible conditions that could

(6)

be used to choose each target class. At the point when the cutoff conditions are settled, the accompanying task is to predict the goal class. In batching, the contemplation isn't to anticipate the target class as all together, it's moreover endeavoring to amass the near kind of things by considering the most satisfied condition, all the things in a comparable social affair ought to be practically identical and should no two particular get-together things to not be similar Experiment results show the predominance of the proposed strategy concerning forecast exactness of Ensemble order report with the highlights chose by outfit, we need a couple of clinical information to apply this model. The precision can be additionally expanded with more information occurrences.

Fig 1: Architecture diagram

Experimental Results

AI strategy called the danger forecast arrangement for hazard factors for cardiovascular sickness. It tries to improve the prescient precision of cardiopathy hazard with a supposed outfit approach. Affiliated arrangement gives high precision and high adaptability, even in the treatment of unstructured information, contrasted with customary order.

The prescient capacity has been tried and achievability of the cardiovascular numerical model in Heart Failure patients to improve the chance of making the mathematic recipe to identify the likelihood of coronary illness events.

The Cleveland Heart Disease Data found in the UCI AI store includes 14 elements assessed on 303 individuals who have coronary sickness. Individuals had been accumulated into five levels of coronary sickness. The information about the contamination status is in the Heart Disease target instructive assortment. Three data diagrams with 303 discernments on the going with 14 variables. The credits of Cleveland informational collection are age, sex, cp - chest torment. Type (regular angina, atypical angina, non-angina torture, asymptomatic), trestbps laying circulatory strain on affirmation, cholserum cholesterol, fbs fasting glucose, rest ecg resting ECG result, thalch most noteworthy heartbeat refined, old apex - ST misery impelled by training related to rest, inclination of the zenith practice ST Segment, ca - number of fluoroscopy concealed vessels, thal reversible defect and class (crippled/healthy).After incorporate lessening step we go tony seven risk factors: cp, thalch, exang, old apex, incline, ca, thalThe test outcomes show contrasted with regular methodologies approaches, for example, Multi-Layer Perceptron's(MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) in view of the examination the planned analytic framework can productively decide the danger level of coronary illness adequately.

Table 1. Algorithm Classification Report

Algorithm RMSE Precision Recall F1 score Accuracy MLP 0.1410 0.1622 0.5422 0.8556 0.9455

CNN 5.2066 1.0 1.0 1.0 1.0

LSTM 0.0 1.1 1.1 1.0 1.0

Load Dataset Data Pre-

processing

Data segmentation training and test

groups

Data classification (MLP/CNN/LSTM)

Ensemble Classification Result

Data Point Clustering

(7)

Fig. 2. Graphical Representation Algorithm Classification

Conclusion

The test outcomes show contrasted with regular methodologies approaches, for example, Multi-Layer Perceptron's (MLP), Convolutional Neural Network (CNN), Long Short-Term Memory (LSTM) in view of the examination the planned analytic framework can productively decide the danger level of coronary illness adequately.Evaluate the integrated model on each complete data, and the confusion matrices returned. Comparing, it can be found that the integrated model improves the performance of the component models and achieves an accuracy MPL(0.9), CNN(1.0), LSTM(1.0). The proposed EDCNN model has end up being a valuable device in the discovery of coronary illness in clinical experts. An extra phase of highlight choice was proposed to improve precision. The dataset is isolated into a preparation set and a test set, and the preparation informational index is utilized to frame singular classifiers. With the test informational collection, the proficiency of the classifiers is tried. The models would thus be able to be used to help patients and medical care experts around the planet in supporting both worldwide and general wellbeing, especially in non-industrial nations and in asset obliged zones with less heart experts accessible. The presentation has additionally improved by the procedures of highlight choice. The component determination methods have added to the precision of the profound learning calculations. In future, advance man- made consciousness has been intended to join to improve the accuracy further.

References

[1] Ansarullah, S.I., & Kumar, P. (2019). A methodical writing audit on cardiovascular issue ID utilizing information mining and AI strategy. Int. J. Late Technol. Eng., 7(6S), 1009–1015.

[2] Haq, U., Li, J.P., Khan, J., Memon, M.H., Nazir, S., Ahmad, S., Khan, G.A., & Ali, A. (2020). 'Intelligent AI approach for compelling acknowledgment of diabetes in E-medical services utilizing clinical information. Sensors, 20(9), 2649.

[3] Haq, U., Li, J., Memon, M.H., & Memon, M.H., Khan, J., & Marium, S.M. (2019). Heart sickness expectation framework utilizing model of AI and consecutive in reverse choice calculation for highlights choice. In Proc. IEEE fifth Int. Conf. Converg. Technol. (ICT), 1–4.

[4] Mohan, S., Thirumalai, C., & Srivastava, G. (2019). Effective coronary illness expectation utilizing crossover AI methods. IEEE Access, 7, 81542–81554.

[5] Tomov, N.S., & Tomov, S. (2018). On deep neural networks for detecting heart disease. arXiv preprint arXiv:1808.07168.

[6] Zappone, A., Di Renzo, M., and Debbah, M. (2019). Remote organizations plan in the period of profound learning: Model-based, AI-based, or both? IEEE Transactions on Communications, 67(10), 7331-7376.

[7] Tuli, S., Basumatary, N., Gill, S.S., Kahani, M., Arya, R.C., Wander, G.S., & Buyya, R. (2020). HealthFog:

An ensemble deep learning based Smart Healthcare System for Automatic Diagnosis of Heart Diseases in integrated IoT and fog computing environments. Future Generation Computer Systems, 104, 187-200.

(8)

[8] Choi, E., Schuetz, A., Stewart, W.F., & Sun, J. (2016). Utilizing repetitive neural organization models for early recognition of cardiovascular breakdown beginning. Diary of the American Medical Informatics Association, 24(2), 361-370.

[9] Das, R., Turkoglu, I., & Sengur, A. (2009). Viable finding of coronary illness through neural organization gatherings. Master frameworks with applications, 36(4), 7675-7680.

[10] Caliskan, A., & Yuksel, M.E. (2017). Order of coronary conduit sickness informational indexes by utilizing a profound neural organization. The EuroBiotech Journal, 1(4), 271-277.

[11] Hasan, N.I., & Bhattacharjee, A. (2019). Profound Learning Approach to Cardiovascular Disease Classification Employing Modified ECG Signal from Empirical Mode Decomposition. Biomedical Signal Processing and Control, 52, 128-140.

[12] Kwon, J.M., Kim, K.H., Jeon, K.H., & Park, J. (2019). Deep learning for predicting in‐hospital mortality among heart disease patients based on echocardiography. Echocardiography, 36(2), 213-218.

[13] Miaoa, K.H., & Miaoa, J.H. (2018). Coronary Heart Disease Diagnosis utilizing Deep Neural Networks.

Worldwide. Journal of Advanced Computer Science and Applications, 9(10), 1-8.

[14] Diller, G.P., Kempny, A., Babu-Narayan, S.V., Henrichs, M., Brida, M., Uebing, A., & Dimopoulos, K.

(2019). AI calculations assessing guess and controlling treatment in grown-up intrinsic coronary illness:

information from a solitary tertiary community including 10 019 patients. European heart diary, 40(13), 1069-1077.

[15] Vennila, V., & Kannan, A.R. (2019). Hybrid Parallel Linguistic Fuzzy Rules with Canopy MapReduce for Big Data Classification in Cloud. International Journal of Fuzzy Systems (IJFS), 21(3), 809-822.

[16] Balamurugan, P., Devi, M.S., & Sharmila, V. (2018). An energy minimizing score based optimal data gathering in wireless sensor networks. International Journal of Engineering & Technology, 7(2.31), 161- 164.

[17] Balamurugan, P., Devi, M.S., & Sharmila, V. (2018). Detecting malicious nodes using data aggregation protocols in wireless sensor networks. International Journal of Engineering & Technology, 7(1.1), 594-598.

[18] Vennila, V., & Kannan, A.R. (2017). Discretized Support Vector Prediction Classifier for Big Data Computation and Information Sharing in Cloud. Asian Journal of Research in Social Sciences and Humanities, 7(2), 566-584.

[19] Sharmila, V., Arasu, G. T., & Balamurugan, P. (2016). Non-Class Element based Iterative Text Clustering Algorithm for Improved Clustering Accuracy using Semantic Ontology. Asian Journal of Research in Social Sciences and Humanities, 6(cs1), 245-257.

[20] BalaMurugan, P., Ravichandran, T., & Sharmila, V. (2016). Grade and Energy based Data Gathering Protocols in Wireless Sensor Networks. Asian Journal of Research in Social Sciences and Humanities, 6(8), 728-744.

[21] Sharmila, V., Balamurugan, P., Vennila, V., & Savitha, S. (2016). Information Retrieval and Recommendation Framework Using Maximum Matched Pattern Based Topic Models. International Journal of Innovative Research in Engineering Science and Technology.

[22] Vennila, V., & Kannan, A.R. (2016). Symmetric Matrix-based Predictive Classifier for Big Data computation and information sharing in Cloud. Computers & Electrical Engineering, 56, 831-841.