Research Papers

Paper Title: Individual Tiger Identification using Transfer Learning Authors: Shailendra Singh Kathait, Co-Founder and Chief Data Scientist, Valiance Analytics Pvt. Ltd., Noida, Uttar Pradesh Vaibhav Singh, Data Scientist, Valiance Analytics Pvt. Ltd., Noida, Uttar Pradesh Ashish Kumar, Principal Data Scientist, Valiance Analytics Pvt. Ltd.,Noida, Uttar Pradesh Summary: The research paper “Individual Tiger Identification using Transfer Learning” presents a novel approach to classify images of tigers into their respective classes, where each class represents an individual tiger. The study leverages deep learning models, specifically the integration of YOLOv8 and EfficientNetB3 models through transfer learning, to accurately identify individual tigers from a dataset of images collected via motion-activated cameras in a large tiger reserve. The paper outlines the significance of individual animal identification for wildlife preservation, particularly in tracking movements and understanding the population density of tigers in their natural habitats. The existing challenge is the classification of species into individual entities, a task made difficult due to the high similarity between individuals of the same species and the vast differences in environmental conditions in which images are captured. The methodology proposed involves a two-step process. Initially, the YOLOv8 model is employed for object detection to create bounding boxes around tigers in images, despite inaccurately labeling them as zebras. This step is crucial for focusing on the tiger within each image, despite the background complexity. Subsequently, the EfficientNetB3 model, pre-trained on the ImageNet dataset, is fine-tuned for the specific task of tiger identification, using a dataset comprising images of 98 unique tigers. This subset was selected based on a minimum availability of 15 images per tiger, from a larger pool of 192 tigers, to ensure sufficient data for reliable model training. Data augmentation techniques, including rotation, horizontal flipping, width and height shifts, and zooming, were employed to address the issue of class imbalance and enhance the robustness of the model. The paper discusses the importance of data preprocessing and augmentation in detail, emphasizing the need for a standardized image resolution and the creation of a balanced training dataset to improve model accuracy. The results demonstrate the model’s high efficiency, achieving a validation accuracy of over 85% and a test accuracy of 88.49% across 98 different tiger classes. The performance is underscored by the precision and recall values, indicating the model’s reliability in individual tiger identification. The paper also discusses the challenges encountered, such as the variability in feature descriptor counts across different images of the same tiger, and the impact of varying backgrounds on SIFT feature extraction results. To conclude, the paper highlights the potential of deep learning and transfer learning in wildlife conservation efforts, particularly in individual animal identification. It suggests that future research could expand the model’s applicability to different geographical locations and diverse environmental conditions, further enhancing its accuracy and robustness. The ultimate goal is to facilitate real-time monitoring of tigers, contributing to better understanding and sustainable human-wildlife interactions. Download Research Paper

Paper Title: Deep Learning-based Model for Wildlife Species Classification Authors: Shailendra Singh Kathait, Ashish Kumar, Piyush Dhuliya and Ikshu Chauhan Summary: Motion-activated cameras have become ubiquitous in ecological parks and wildlife sanctuaries, capturing images upon sensor-triggered motion, including infrared visuals that were once impractical. Despite this technological leap, extracting pertinent wildlife information from the vast image dataset remains time and labor-intensive. This paper presents a groundbreaking solution, employing deep learning models, particularly the VGG16 ConvNet architecture through transfer learning, to achieve near-human-level accuracy in information extraction. The study focuses on a dataset of 33,511 images representing 19 species from the Ladakh region of India. Training and testing the model yielded an impressive accuracy of 89.12%. The established pipeline exhibits vast potential for wildlife monitoring in various national parks, advancing ecological research and conservation. The methodology involved utilizing 80% of the dataset for training and 20% for validation. Subsequently, 3309 unseen images were tested, leading to a confusion matrix. The matrix highlights accurate species classifications, such as correctly identifying 284 out of 302 bird images. However, the study acknowledges geographical limitations, emphasizing the need for a region-specific model. Enhancements in overall test accuracy are anticipated through increased and diverse training data, optimizing the model’s efficiency beyond the Ladakh region. Download Research Paper

Paper Title: Custom Deep Neural Network For Message Scoring Authors: Shailendra Singh Kathait, Ashish Kumar and Ikshu Chauhan Summary: This paper offers customized message scoring creation of Technical Literature of Drugs using Text Analytics and Deep Learning. A customized Deep Learning driven model has beendeveloped that feeds scientific data (texts) about medicine(s) (written by medical researchers) & creates multiple sub messages that can be used for marketing research to effectively communicate benefits of drugs. Model helped in reducing manual intervention by automating the complete process and replacing Linguists. Download Research Paper

Paper Title: Application of Image Processing and Convolution Networks in Intelligent Character Recognition for Digitized Forms Processing Authors: Shailendra Singh Kathait, Shubhrita Tiwari Summary: Image processing is a rapidly evolving field with immense significance in science and engineering. One of the latest applications of Image processing is in Intelligent Character Recognition (ICR). Intelligent Character Recognition is the computer translation of handwritten text into machine-readable and machine-editable characters. It is an advanced version of Optical Character Recognition system that allows fonts and different styles of handwriting to be recognized during processing with high accuracy and speed. ICR, in combination with OCR and OMR (Optical Mark Recognition), is used in forms processing. Forms processing is a process by which one can capture information entered into different data fields filled in forms and convert it to an editable text. Forms processing systems can range from the processing of small application forms to large scale survey forms with multiple pages. The Recognition Engine, designed using Image Processing and Convolution Networks helps save time, labor and money in addition to the increase of accuracy. Download Research Paper

Paper Title: Intelligent Recommendation Engine Authors: Shailendra Singh Kathait, Shubhrita Tiwari, Piyush Kumar Singh Summary: Highly engaged visitor is central to the success of any Digital publisher. Having right content and serving it to the right audience at right time is the key to achieving high engagement rates. Publishers have used machine learning based recommendation engines to summarise articles, extract article metadata (keywords, phrases), build user preferences and recommend right content to the visitor in real time on a website, as email digest or app notifications. This has been achieved through collaborative filtering, content-based filtering. Recommendation algorithms based only alone on collaborative filtering or content based filtering overlook important factors that drive suitable recommendations for a user. This paper describes the application of hybrid approach i.e. Collaborative Filtering together with Content-based Filtering for making the most appropriate and relevant recommendations. Download Research Paper

Paper Title: Supplier Evaluation Model on SAP ERP Application using Machine Learning Algorithms Authors: Manu Kohli Summary: For business enterprises, evaluating a supplier is a mission critical process. On ERP (Enterprise Resource Planning) applications such as SAP, the supplier evaluation process is performed by configuring a linear score model, however this approach has a limited success. Therefore, author in this paper has proposed a two-stage supplier evaluation model by integrating data from SAP application and ML algorithms. In the first stage, author has applied data extraction algorithm on SAP application to build a data model comprising of relevant features. In the second stage, each instance in the data model is classified, on a rank of 1 to 6, based on the supplier performance measurements such as on-time, on quality and as promised quantity features. Thereafter, author has applied various machine learning algorithms on training sample with multi-classification objective to allow algorithm to learn supplier ranking classification. The application of supplier evaluation model proposed in the paper can be generalised to any other other information management system, not only limited to SAP, that manages Procure to Pay process. Download Research Paper

Paper Title: Unsupervised Key-phrase Extraction using Noun Phrases Authors: Shailendra Singh Kathait, Shubhrita Tiwari, Anubha Varshney, Ajit Sharma Summary: Increasing volume of digital content in the form of articles, blogs has accelerated the need for an automated process can simplify the extraction of relevant tags and even summarize the content on an article. Traditionally supervised machine learning approaches have been used but their utility needs labeled datasets which become a bottleneck in learning as content grows exponentially. Facing similar scenario we have experimented with an unsupervised approach to key-phrase extraction that doesn’t depend on labeled training datasets. Details are outlined in the research paper. Download Research Paper

Paper Title: Hybrid Intelligent System via Fuzzy Regression Analysis, Bayesian Gaussian Reasoning Model in Healthcare Authors: Shailendra Singh Kathait, Dr Aankita Kaur, Anubha Varshney Summary: In this paper, we propose the architecture of Hybrid Intelligent System with different techniques of pattern recognition and machine learning. Fuzzy Regression and Bayesian Gaussian Neural Network approach are used to build the model. Fuzzy regression deals with the uncertain, vagueness of the system. Naive Bayesian classifier helps in building strong independent relationships whereas Gaussian classifiers correlates high dimensional data with kernel function to yield better performance of the system. A hybridized combined approach of neural network is presented in healthcare. It is due to its flexibility of modeling, and robust nature, learning ability from complex functions and the application of different algorithm for reduction of errors for a better intelligent system. Download Research Paper

Paper Title: Integrating Neural Networks with Time Series Forecasting: Improving Sales Authors: Shailendra Singh Kathait, Dr Aankita Kaur, Shubhrita Tiwari, Anubha Varshney Summary: This paper presents a comparison between different algorithms that are used for time series forecasting on a noisy time series data. The accuracy of forecasting is first evaluated using traditional methods like Moving Average, ARIMA modeling etc. and then Machine Learning based approach with ANN (Artificial Neural Networks) is used for forecasting. ANN algorithms have found to deliver superior & robust predictions. Download Research Paper

Paper Title: Predicting Equipment Failure On SAP ERP Application Using Machine Learning Algorithms Authors: Manu Kohli Summary: A framework model to predict equipment failure has been keenly sought by asset intensive organisations. Timely prediction of equipment failure reduces direct and indirect costs, unexpected equipment shut-downs, accidents, and unwarranted emission risk. In this paper, the author has proposed an equipment reliability model, for equipment type pumps, designed by applying data extraction algorithm on equipment maintenance records residing in SAP application. Author has initially applied unsupervised learning technique of clustering and performed classes to cluster evaluation to ensure generalisation of the model. Thereafter as part of supervised learning, data from the finalised data model was fed into various Machine Learning (ML) algorithms where the classifier was trained, with an objective to predict equipment breakdown. The classifier was tested on test data sets where it was observed that support vector machine (SVM) and Decision Tree (DT) algorithms were able to classify and predict equipment breakdown with high accuracy and true positive rate (TPR) of more than 95 percent. Download Research Paper