Info
Berufserfahrung und Ausbildung
Veröffentlichungen
-
Short Term Ambient Temperature Forecasting for Smart Heaters
IEEE
There is an complex relationship between power dissipation at edge by heaters and ambient temperature in zones. We abstract the dynamics of interaction by learning a sequence to sequence forecasting model. Our methodology gives near to real time prediction at a resolution of 1 minutes with an error margin of less than 2 degree Celsius. This is instrumental in modelling a smart heater’s thermal demand response in real life deployment.
-
Recommendation system based on product purchase analysis
Innovations in Systems and Software Engineering, a NASA journal
A new type of recommendation system proposed as well as Amazon's network analyzed to get deep insights.
Andere Autor:innenVeröffentlichung anzeigen -
Economical Analysis of Flexibility in Smart Grids
As energy demand increased and production means diversified, conventional approaches of looking into distribution grids need to evolve. The Smart Grid paradigm introduces new possibilities of real-time market sensing and interaction models between producers and consumers. In particular, by understanding the types of consumers and their potential willingness to adapt their energy demand with price incentives, innovative pricing strategies in the Smart Grid are expected to lead to better…
As energy demand increased and production means diversified, conventional approaches of looking into distribution grids need to evolve. The Smart Grid paradigm introduces new possibilities of real-time market sensing and interaction models between producers and consumers. In particular, by understanding the types of consumers and their potential willingness to adapt their energy demand with price incentives, innovative pricing strategies in the Smart Grid are expected to lead to better production management, profit maximization and end consumers satisfaction levels. In this work we propose a novel framework and a simulation scenario of a global energy network with heterogeneous types of producers and consumers from which different types of behaviors and interactions can be studied.
Andere Autor:innenVeröffentlichung anzeigen -
Greedy Supervised Feature Selection (GSFS) of Physicochemical Properties of Amino Acids
International Journal of Innovative Research in Science, Engineering and Technology
Feature selection is an immensely important task in the domain of pattern recognition. Physicochemical properties of amino acids have been extensively used in classification and regression tasks like protein secondary structure prediction, protein post translational modification etc. Out of the 544 features, presented in AAINDEX, not all are relevant for a certain classification. Classifiers like Neural Network (NN) or Support Vector Machine (SVM) will include information from all the features…
Feature selection is an immensely important task in the domain of pattern recognition. Physicochemical properties of amino acids have been extensively used in classification and regression tasks like protein secondary structure prediction, protein post translational modification etc. Out of the 544 features, presented in AAINDEX, not all are relevant for a certain classification. Classifiers like Neural Network (NN) or Support Vector Machine (SVM) will include information from all the features in order to model a problem. Hence, presenting the most appropriate features to a classifier will enhance its performance. Feature selection algorithms attempt to select a subset of the available features which will help perform classification better. Feature selection methods can be supervised or unsupervised. In this paper we present a greedy supervised algorithm (GSFS) to select the most apt features for a particular classification task. We have implemented our algorithm on the problem of protein phosphorylation of human proteins and have obtained better results than other works which have implemented an unsupervised algorithm for the same task. We have achieved an increase of area under roc curve of about 1% in 5 fold cross validation experiments.
Andere Autor:innen
Projekte
Auszeichnungen/Preise
-
Persyval scholarship
Persyval Labs
Granted for academic excellence during postgraduate studies. 10/230 applicants were selected in total from multiple disciplines.
-
State Science Fare Winners
Government of West Bengal, India
Develop an aerial unmanned solution that can detect people trapped in disaster-hit places and beacon spatial coordinates for aero-dropping relief aid. We also tested a swarm of drones working in unison to parallely cover geospatial patches.
Ebenfalls angesehen
Weitere Mitglieder, die Angan M. heißen
Es gibt auf LinkedIn 0 weitere Personen, die Angan M. heißen.
Weitere Mitglieder anzeigen, die Angan M. heißen