Hybrid LCNN for Beneficial Student Career Prediction: Leveraging the Potential of LSTM and Optimized CNN

Abstract

Student career prediction has been done recently by measuring the variables that might affect career decisions with the use of questionnaires or diagnostics. However, it is challenging to accurately predict someone's professional choices because of the complexity of each person's beliefs and ambitions. This paper uses behavioural data from students to forecast career options to overcome this challenge. Rigorous pre-processing procedures are used to clean and validate initially collected data. Then the feature extraction includes central tendency measures, quantitative measures of dispersion, and Mean Absolute Deviation (MAD) to characterize the dataset's distribution and variability. After feature extraction, introduce a new Hybrid LCNN method, which combines long-term and short-term memory (LSTM) with optimized Convolutional Neural Network (CNN) architecture. This optimized CNN and LSTM get extracted features in parallel alongside finely tuned time model training. Then introducing the Barnacle Pathfinder Optimization Algorithm (BPOA) which combines the concepts of Barnacles Mating Optimizer (BMO) and Pathfinder Algorithm (PFA), hyperparameter optimization of LSTM and CNN Also, use BPOA models to solve CNN weights, and effect on the final forecast result Increase it. The predictions produced by the LSTM and optimized CNN components are combined using a weighted mean to exploit the synergistic effects of the two architectures. The proposed Hybrid LCNN model achieved the accuracy of 99.25% for learning rate 70% and 99.48% for learning rate 80%.