Browsing by Author "Islam, Rifatul"

Now showing 1 - 2 of 2

An approach towards a sustainable urban city - utilization of existing rooftop solar energy panels by making use of DC appliances
(BRAC University, 2021-06) Rifat, Ashir Hamim; Hossain, Muckbul; Islam, Rifatul; Noman, Shoud; Azad, A. K. M. Abdul Malek
As with the growing GDP, developing market economy of Bangladesh is raising question on energy sustainability. Although having the infinite possibilities of harnessing solar energy, it could not be utilized properly as 90% of the urban rooftop solar panels are idle. The proposed system can work with the existing rooftop solar panels by using them as the source for powering the DC appliances of the apartments and store the excess energy in the 48V battery bank for later use. National grid connection is also kept as a backup source for powering the appliances when solar irradiance is unavailable and the batteries are all drained up. The reason of being exceptional of this project is use of DC appliances instead of converting the DC solar energy into AC which is inefficient and at the same time costly. This project successfully shows that those idle rooftop solar panels can utilized in the most effective and efficient ways while saving money and contributing to the nation’s energy sustainability.
Predicting effectiveness of marketing through analyzing emotional context in advertisement using deep learning
(BRAC University, 2020-04) Arafat, Sheikh Mohammad; Islam, Rifatul; Rafi, Ishraque Arefin; Islam, Md. Rashedul; Alam, Md. Golam Rabiul
In this modern age, marketing strategy is becoming a new challenge. Not only the global market but also people’s choices are shifting to catch the attention of buyers. Also, based on consumer’s choice organizations are bringing changes in their marketing policy to increase the chances of their product selling rate. Basically, to promote their products and grab buyer’s attention they are promoting advertisements on every media platform. But they are not aware of the effectiveness of marketing and which emotional states are needed more and which are not needed much. Therefore, we lead this study to recognize a successful advertisement and identify the rate of the emotional states which make good impact in people mind to purchase the product. Using deep learning and supervised machine learning algorithms as well as feature extraction methods for instance, LSTM-RNN, SVM, XGBOOST, Na¨ıve Bayes, Multiple Linear Regression, MFCC, Zero-Crossing Rate, Power Spectral Density, we find out and evaluate the rate of the emotional states to figure out the liking and purchase intent which makes an advertisement successful.