Rajiv Gandhi College of Engineering

Stock price prediction remains challenging because financial markets show extensive volatility as well as nonlinearity. We have developed ResNLS which combines the Residual Networks (ResNet) and Long Short-Term Memory (LSTM) networks for superior predictive accuracy at stock price prediction. Through the ResNet component deep hierarchical features are extracted and the vanishing gradient problem is reduced while the LSTM network identifies difficult temporal relationships to improve analytical trend recognition. Our innovative hybrid model design utilizes residual learning together with sequential modelling thus producing more effective outcomes than regular deep learning systems. The accuracy of ResNLS reaches 94.3% during real-world stock market evaluations beyond traditional models with reduced error rates. The developed methodology solves extraction issues together with time dependencies while providing an effective solution for predicting financial time series.

This study reports the findings of an experimental exploration for application of ground granulated blast furnace slag (GGBS) as a substitute for cement in various percentages. The research focuses on evaluating the effects of varying GGBS dosages, ranging from 0

Maximum power point tracking (MPPT) circuits are developing presently to generate switch signals with appropriate duty cycles for switches to enable photovoltaic (PV) systems to operate at maximum output power typically through conventional boost converters. This primary challenge, especially for Microgrids and small scale energy-power systems, the voltage provided by the boost converter (i.e., MPPT circuit) is insufficient to enhance the resulting maximum power point voltage. This work introduced a novel high gain for voltage of a DC-DC boost converter and designed to work along with conventional MPPT converter. A switched capacitor unit is implemented at the output side of the device to increase the gain of voltage and minimize its stress on the electronic switches. Hence, this can be a dynamic aspect for the longevity of both the PV panel. The components of the proposed converter, especially semiconductor devices and the performance of the converter are evaluated while considering raid changes in irradiation as well as temperature caused by fluctuating weather conditions. A prototype device at a workshop is employed and analyzed the results to verify the conclusions of theoretical and simulation studies.

Traditional lock systems and single-factor electronic security methods are vulnerable to unauthorized access, key duplication, and password theft. To overcome these limitations, this project presents a secure multi-factor smart locker system using RFID authentication, PIN verification, and fingerprint recognition. The system is developed using an Arduino Uno microcontroller integrated with an RC522 RFID reader, 4×4 matrix keypad, R307 fingerprint sensor, LCD display, relay module, buzzer, and solenoid door lock. The authentication process is performed in three stages. First, the user scans an RFID card. If the card is recognized, the system requests a 4-digit PIN through the keypad. After successful PIN verification, fingerprint authentication is performed. Only when all three stages are verified successfully does the relay activate the solenoid lock to open the door. For additional security, the system triggers a buzzer alarm after three incorrect attempts and temporarily locks the system to prevent unauthorized access. The proposed smart locker system provides higher security, reliability, and faster response compared to traditional locking systems, making it suitable for homes, offices, banks, laboratories, and other restricted areas.