The "INDIAai" (National AI Portal of India) portal provides comprehensive articles every week highlighting the research contributions made by universities in India.

Our objective is to offer thorough reporting on the AI research contributions made by a single university every week. This series allows researchers and students to provide concise explanations of their work.

Top AI research contributions from Amrita Vishwa Vidyapeetham Kochi Campus

A multivariate EMD-LSTM model aided with Time-Dependent Intrinsic Cross-Correlation for monthly rainfall prediction.

Kavya Johny, Maya L. Pai, and Adarsh S are the authors of this research. The authors say, "Meteorologists worry about predicting non-linear systems like rainfall time series due to their many regulating components, complicated interrelationships, and multiscaling behaviour. Hence, hybrid modelling involving decomposition techniques is preferable over strenuous physical models and standalone-driven methods for accurate rainfall predictions." 

They further said, "This study proposes a novel hybrid modelling framework integrating Long Short Term Memory (LSTM) and Multivariate Empirical Mode Decomposition (MEMD) aided with Time-Dependent Intrinsic Cross-Correlation (TDICC) analysis algorithm for monthly rainfall predictions. Over the years, extreme weather events have increased due to global warming and climate change. The proposed model is significant for the extremes like drought and flood management. It is an essential task to be executed for the proper risk management and disaster preparedness in countries like India thriving on the agro-based economy."

Evaluating the performance of bagging-based k-nearest neighbour ensembles with the voting rule selection method.

M. S. Suchithra and Maya L. Pai contributed to this research work together. The authors say, "The key field of machine learning is preference learning because of triggering preference models from the available information. In recent years, the research interest for considering label ranked data has increased from a data mining viewpoint, and there is a genuine area of preference modelling research recognized by recommendation systems." 

They continued, "This work includes exploring label ranked data based on the popular label ranking algorithms to identify the best-performing model under the defined parameter settings. A novel approach is introduced by integrating the bagged k-nearest neighbour model with the Voting Rule Selection (VRS) method. It is designed to evaluate the performance of the bagged k-nearest neighbour model with VRS."

Improving miRNA Disease Association Prediction Accuracy Using Integrated Similarity Information and Deep Autoencoders

Sujamol S, Vimina E R, and U Krishnakumar jointly contributed to this research work. The authors state, "Many biological and molecular functions in the body depend on miRNAs. Many complicated illnesses are linked to miRNA dysregulation. This research focuses on developing AI-assisted methodology for predicting unknown miRNA-disease associations."

Topology Analysis of Protein-Protein Interactome for Prioritizing Key Comorbid Genes via Sub Graph Average Path Length Centrality.

Nikhila T. Suresh, Vimina E. R., and U. Krishnakumar jointly contributed to this research work. The researchers express that "Comorbidity refers to the co-occurrence of multiple diseases in the same individual. Diseases sharing common molecular mechanisms or entities produce comorbid states and have notable indications over the disease management and the progression." 

They further described, "This work focuses on prioritizing the comorbid genes via Protein-Protein Interaction Network (PPIN) analysis, which can be used to identify putative disease biomarkers that can be repurposed for the management of comorbidity."

Protein Secondary Structure Prediction Using Cascaded Feature Learning Model

Geethu S and Vimina E.R contributed to this research work. The authors say, "Proteins are macromolecules responsible for numerous biological processes in a living organism. Awareness of the protein structure can shed light on understanding the organism's functions and other molecular mechanisms, aid in drug design and development, etc." Furthermore, they stated, "This research focuses on developing a Deep Learning methodology for protein secondary structure prediction."