Published 5 papers in International Journals like Springer, IEEE, Journal of physics
2 more papers under review
Interests: Machine Learning, Data Analytics, Deep Learning, Internet of Things,
Domains: Healthcare, Nature, Society, Finance
Early prediction model for coronary heart disease using genetic algorithms, hyper-parameter optimization and machine learning techniques
Coronary Heart Disease (CHD) is one of the major causes of morbidity and mortality worldwide. According to the World Health Organization (WHO) survey, Cardiac arrest accounts for more deaths annually than any other cause. But the silver lining over here is that heart related diseases are highly preventable, if simple lifestyle modifications are carried out. However, it is a challenging factor to identify high risk heart patients at times due to other comorbidity factors such as diabetes, high blood pressure, high cholesterol and so on. Hence it is needed to develop an efficient early prediction model which can detect high risk patients and their life could be saved.The proposed system helps to identify the best set of features for diagnosis using traditional machine learning algorithms along with modern Gradient Boosting approaches. Genetic algorithm for feature selection to optimize performance by reducing the number of parameters by 20% whilst keeping the accuracy of the model intact is implemented in the proposed system. In addition, hyper parameter optimization techniques are executed to further improve the predictive model’s performance.
Hybrid Feature Selection and Bayesian Optimization with Machine Learning for Breast Cancer Prediction
Breast Cancer is one of the most ubiquitous type of cancer among women and rarely found in men. According to the World Health Organization (WHO), Cancer is defined as an uncontrollable abnormal growth of cells in any organ or tissue of the body. Neoplasm or Malignant tumor are common words that describe cancer. According to a survey conducted by the World Health Organization (WHO), cancer accounts for deaths of approximately 9.6 million people globally, and is the second most prevalent cause of death in humans. Cancer is responsible for one out of every six deaths across the world. The good thing is cancer if diagnosed at an early stage, is likely to be treated successfully resulting in high survival rate and low treatment cost. However, it is quite difficult to diagnose it at an early-stage. Therefore, there is a necessity for an efficient cancer prediction model that can predict breast cancer at an early stage. The proposed model highlights the finest set of features required for the detection of breast cancer. It uses Bayesian optimization technique along with hyper parameter tuning and feature selection techniques to decrease the number of parameters by almost 40% while maintaining high accuracy. The best accuracy of 96.2% is obtained with Extra tree classifier algorithm by using feature selection technique along with bayesian optimization and hyperparameter tuning.
Ensemble Methods With Bidirectional
Feature Elimination For Prediction And
Analysis Of Employee Attrition Rate During
In the wake of the COVID-19 pandemic, a myriad of organizations across the globe have decided to churn out some of their work-force owing to the economic recession. According to the International Monetary Fund (IMF), the world has experienced a financial shrink- age of 3% which is the steepest slowdown since the Great Depres- sion in the 1930s. Aviation, Tourism, Travel, and Hospitality are the industrial sectors that have been impacted the worst. Deloitte In- dia’s 2020 Workforce and Increment Trends Survey has stated that the involuntary attrition rate for the current financial year is close to 15%. The research conducted aims at analyzing the Attrition of employees based on factors like their Educational Qualifications, Years of Work Experience, Gender, Department, and many others. The proposed system also predicts the Attrition rate of employees using a Machine Learning Pipeline that uses Advanced Ensem- bling, Gradient Boosting, Feature Selection through Bi-Directional Elimination, and optimizing the hyper-parameters through a Ran- domized Grid Search Approach. Owing to the optimizations carried out in the entire Model Building pipeline, the algorithm successfully achieves state-of-the-art performance. To ensure legitimacy in the results, the Stratified K Fold Cross Validation methodology is used for evaluation.
Suvarga: Promoting a Healthy Society
In India, over 22% of the population is below the poverty line. This poverty pushes people on streets which in the future transforms into slums. These slums, as are not planned, lack certain necessities like electricity, sanitary services, and basic hygiene resources leading to a hub for the spread of diseases. In essence, the primary aim of this paper is to identify the leading causes of diseases in slum areas of Mumbai using data collected from IoT modules, health checkup drives, and various government authorities. With this information, the concerned civic authorities and slum residents will be alerted regarding the danger so that necessary action can be taken. This, in turn, promotes the healthier society in various slum regions of India.
Patang Abhidhani - Convolution Neural Network based Butterfly Research Survey
There has been a great loss of biodiversity worldwide and efforts have to be taken in order to restore it. To restore biodiversity, it's crucial to know the cause for its decline and the role played by the insects in it. In India, butterflies are not given much importance in the conservation species. Since there are no names for butterflies in regional languages, the only access for comprehensive information for people's awareness is really nothing. This paper aims at providing details of research done on different butterfly species, their existence once in different parts of the world, and technologies employed for solving the problems like identification of different species, distribution, restoration, and rejuvenation of diversity.