Article Preview
TopIntroduction
An optimization problem refers to finding a set of parameter values in the feasible solution set so that the objective value of the problem can reach the maximum or minimum under some constraints. Various optimization problems in life have prompted the progress of algorithm research. Intelligent optimization algorithms have been vigorously developed in this regard. At present, the utilization of information flow in the network has become a hot topic, and the research on sentiment analysis of large-scale data in the network has gradually attracted the attention of scholars. Nowadays, the most commonly used sentiment analysis method is the text sentiment analysis method based on deep learning. However, due to the large number of hyperparameters involved in this method, an appropriate optimization method is needed to improve the efficiency of this sentiment analysis method.
A Particle Swarm Optimization algorithm (PSO) is a class of classical swarm intelligence optimization algorithms (Kennedy & Eberhart, 1995). A PSO algorithm has a simple principle and fast convergence speed. It has undergone extensive research and development in recent years in many fields, such as resource transportation (Singh & Singh, 2023), system reliability tests (Li et al., 2022), high-speed train head shape optimization (He et al., 2022), etc. However, the PSO algorithm, like other swarm intelligence optimization methods, tends to premature maturity and falls into the local optimum (Ding & Gu, 2020). For this reason, many PSO variants have been proposed to enhance the search performance of a PSO. Common variants include using parameter adaptive control to optimize the execution effect of the algorithm, mixing a PSO algorithm with other algorithms or strategies to improve the performance of the algorithm, using different neighborhood structure to optimize the search ability of the algorithm, using multiple population mechanisms to optimize the efficiency of population information interaction, and changing the learning mechanism of particles to improve the performance.
The PSO variants proposed above will be optimized for different characteristics according to their respective algorithm strategies. They are highly dependent on the problem and are not suitable for solving sentiment classification problems. In this study, a hybrid learning particle swarm optimization with a fuzzy logic (HLPSO-FL) algorithm is proposed for sentiment analysis in the context of deep learning. The main contributions of this paper are as follows:
- 1.
Based on the scale-free network topology, a mainstream learning strategy is proposed to reduce the speed of algorithm information transmission and avoid the algorithm falling into the local optimum.
- 2.
Utilizing a random learning method as opposed to the self-learning strategy enhances the population's diversity and prevents its early convergence.
- 3.
According to the state of each particle, fuzzy logic is introduced to dynamically control parameters, such as inertia weight and individual learning factor of each particle, and dynamically adjust the exploration and development capabilities.
The rest of the paper is structured as follows: The related work of particle swarm optimization algorithm, scale-free network, and sentiment analysis approach is introduced in Related Works. Hybrid learning particle swarm optimization with fuzzy logic details the algorithms outlined in this study. The experimental analysis is arranged in Experimental Results and Analysis. Finally, the summary and prospect of the work of this paper are given.