Kailash Kumar Baraskar

@medicaps.ac.in

Assistant Professor Department of Computer Science and Engineering Medicaps University Indore
Medicaps University Indore

Kailash Kumar Baraskar


                  Download Compact PDF  
https://researchid.co/kailash2204

RESEARCH, TEACHING, or OTHER INTERESTS

Artificial Intelligence, Computer Science, Computer Science Applications, Automotive Engineering
6

Scopus Publications

Scopus Publications

  • ResNet-152 for Early Skin Cancer Detection: A Deep Learning Approach to Medical Imaging
    16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
  • Energy Consumption Prediction using Ensemble Learning and Streamlit Application
    Priyanka Anup Ujjainkar, Tabassum H Khan, Smita Nirkhi, Sanskar Mudpalliwar, Shivang Chavan, Kailash Kumar Baraskar
    2025 International Conference on Artificial Intelligence and Quantum Computation Based Sensor Application Icaiqsa 2025 Proceedings, 2025
    This project introduces a user-friendly web application for predicting household energy usage using ensemble learning techniques. Built with Streamlit, the tool enables users to upload their own energy datasets, explore usage patterns, and receive real-time predictions with just a few clicks. To ensure accurate forecasting, the system includes essential preprocessing steps such as managing missing values, encoding categorical variables, and extracting relevant date and time features. The ensemble model used in the application delivers strong predictive performance, achieving an overall accuracy of 95.8%. It also maintains a Mean Absolute Error (MAE) of 12.5 kWh and a Root Mean Squared Error (RMSE) that falls within acceptable limits. These results indicate the model is well-suited for real-world applications in household energy forecasting. Feature analysis showed that the most influential factors affecting energy consumption were the number of appliances, average daily usage hours, and occupancy levels—factors that reflect common trends in actual energy use. In addition to delivering predictions, the system offers automated suggestions to help users improve model performance. These include collecting more historical usage data, enhancing feature engineering methods, and experimenting with hyperparameter tuning. The platform’s interface is simple and intuitive, making it suitable for both technical users and those without a background in data science. The overall findings suggest that ensemble learning can play a valuable role in promoting smarter and more efficient energy use at the household level. Looking ahead, future improvements may include incorporating real-time IoT data, developing more advanced hybrid models, and expanding the tool’s capabilities to support time-series forecasting for improved adaptability.
  • Deep Learning Models for Automated Detection and classification of Fungal and Bacterial Infections in Agriculture Crop
    Rina Kailas Parteki, Hrushikesh Madhukar Panchabudhe, Kailash Kumar Baraskar, Nitin Chakole, Mayuri Abhijit Getme, Praveen Kumar Dhankar
    2025 1st International Conference on Advancement in Futuristic Technologies Icaft 2025, 2025
    It is relevant to find and categorize bacterial and fungal diseases in crops to ensure that there is sufficient food, minimize loss of crops and enable farmers to continue growing in the long run. The conventional methods of identifying the diseases affecting plants are labor-intensive, costly in time, and subject to human error. This is an indication of the significance of having effective and secure solutions. This research examines how the challenge of automated detection and classification of infections can be addressed using deep learning models using big picture data sets on publicly available archives and actual farms. The proposed system involves noise suppression, normalization and improvement as pre-processing measures to ensure that datasets are more varied and trustworthy. Some of the tested deep learning designs include the Convolutional Neural Networks (CNNs), ResNet, DenseNet, and EfficientNet. They both have advantages as regards feature extraction and hierarchical learning. Also added are hybrid and ensemble-based methods to enhance the capability of classifying infections into more than one category. The mathematicians consider picture classification as an issue that has two or more solutions. The suggested solution not only is highly effective in experimentation, but it can also be applied to smart farming systems immediately as it is scalable. Ultimately, this research will enhance precision agriculture by providing farmers and agronomists with an effective tool in the rapid and accurate decision-making regarding the way to cope with diseases.
  • Visualizing Dependencies in Go
    15th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2024, 2024
  • Deep Learning Based IDS to Detect Anomaly over Social Networking Site: Comprehensive Review
    Safdar Sardar Khan, Arpit Deo, Kailash Kumar Baraskar, Amit Patel, Amritansh Pathak, Aditya Kumar Joshi
    2023 International Conference on Integration of Computational Intelligent System Icicis 2023, 2023
    One of the primary purposes of an Intrusion detection system (IDS) is to categorize the regular and irregular activities on the online social network (OSN). The complexity of patterns of communication in OSN makes the intrusion detection a challenging task as compared to traditional networks. Recent work aims at soft computing technique based intrusion detection. This paper is an analytical study of emerging deep learning techniques which will assist the researchers to identify normal and abnormal behavior of users on any social networking site. We propose to use the Decision Tree Classifier followed by histogram of oriented gradients (HOG) respectively for feature extraction and representation followed by training on a deep Convolution Network to detect the presence of anomalies. Performance measures, generic datasets and hybrid techniques are also suggested to evaluate deep learning approaches in IDS for OSN.
  • Recognising Human Actions Using Long-Term Recurrent Convolutional Network (LRCN)
    Safdar Sardar Khan, Arpit Deo, Kailash Kumar Baraskar, Aadity Gangrade, Aman Verma, Ansh Jain
    3rd IEEE International Conference on ICT in Business Industry and Government Ictbig 2023, 2023
    The study of human activity recognition has gained popularity. It is employed to make sense of the actions taken by the people in the films. The CNN and LSTM models, which were trained independently, can be used to do this. Using a pre-trained model, we can use this CNN model to extract spatial characteristics from the frames that were taken from the movies. The action depicted in the videos can then be predicted using an LSTM model by using the features that CNN extracted. Alternatively, combining convolutional and LSTM layers into a single model using the Long Recurrent Convolutional Network is a more effective way to implement it. We discovered that using a single model was more accurate than using each model separately.