International Journal of Computational Intelligence and Applications

The identification of opinions and sentiments from tweets is termed as “Twitter Sentiment Analysis (TSA)”. The major process of TSA is to determine the sentiment or polarity of the tweet and then classifying them into a negative or positive tweet. There are several methods introduced for carrying out TSA, however, it remains to be challenging due to slang words, modern accents, grammatical and spelling mistakes, and other issues that could not be solved by existing techniques. This work develops a novel customized BERT-oriented sentiment classification that encompasses two main phases: pre-processing and tokenization, and a “Customized Bidirectional Encoder Representations from Transformers (BERT)”-based classification. At first, the gathered raw tweets are pre-processed under stop-word removal, stemming and blank space removal. After pre-processing, the semantic words are obtained, from which the meaningful words (tokens) are extracted in the tokenization phase. Consequently, these extracted tokens are classified via optimized BERT, where biases and weight are tuned optimally by Particle-Assisted Circle Updating Position (PA-CUP). Moreover, the maximal sequence length of the BERT encoder is updated using standard PA-CUP. Finally, the performance analysis is carried out to substantiate the enhancement of the proposed model.

Although distance measures are used in many machine learning algorithms, the literature on the context-independent selection and evaluation of distance measures is limited in the sense that prior knowledge is used. In cluster analysis, current studies evaluate the choice of distance measure after applying unsupervised methods based on error probabilities, implicitly setting the goal of reproducing predefined partitions in data. Such studies use clusters of data that are often based on the context of the data as well as the custom goal of the specific study. Depending on the data context, different properties for distance distributions are judged to be relevant for appropriate distance selection. However, if cluster analysis is based on the task of finding similar partitions of data, then the intrapartition distances should be smaller than the interpartition distances. By systematically investigating this specification using distribution analysis through the mirrored-density (MD plot), it is shown that multimodal distance distributions are preferable in cluster analysis. As a consequence, it is advantageous to model distance distributions with Gaussian mixtures prior to the evaluation phase of unsupervised methods. Experiments are performed on several artificial datasets and natural datasets for the task of clustering.

With the continuous improvement of science and technology, the substation remote control system has been constantly improved, which provides the possibility for the complete realization of intelligent and unmanned substation. However, due to the special substation environment, it is easy to cause interference, coupled with the low accuracy of today’s video image processing algorithm, which leads to the frequent occurrence of false alarms and missing alarms. Manual intervention is needed to deal with this, which inhibits the display of automatic intelligent substation processing functions. Therefore, in this paper, the most rapidly developed machine learning algorithm — deep learning is applied to the substation instrument equipment identification processing, in order to improve the accuracy and efficiency of instrument equipment identification, and make due contributions to the full realization of unattended substation.

Environmental sound classification (ESC) is a challenging problem due to the complexity of sounds. To date, a variety of signal processing and machine learning techniques have been applied to ESC task, including matrix factorization, dictionary learning, wavelet filterbanks and deep neural networks. It is observed that features extracted from deeper networks tend to achieve higher performance than those extracted from shallow networks. However, in ESC task, only the deep convolutional neural networks (CNNs) which contain several layers are used and the residual networks are ignored, which lead to degradation in the performance. Meanwhile, a possible explanation for the limited exploration of CNNs and the difficulty to improve on simpler models is the relative scarcity of labeled data for ESC. In this paper, a residual network called EnvResNet for the ESC task is proposed. In addition, we propose to use audio data augmentation to overcome the problem of data scarcity. The experiments will be performed on the ESC-50 database. Combined with data augmentation, the proposed model outperforms baseline implementations relying on mel-frequency cepstral coefficients and achieves results comparable to other state-of-the-art approaches in terms of classification accuracy.

Products sorting is a task of paramount importance for many countries’ agricultural industry. An accurate quality check assures that good products are not wasted, and rotten, broken and bent food are properly discarded, which is extremely important for food production chains. Such products sorting and quality controls are often performed with consolidated instruments, since simple systems are easier to maintain, validate, and they speed up the processing in terms of production line speed and products per second. Moreover, industries often lack advanced formation, required for more sophisticated solutions. As a result, the sorting task for many food products is mainly done by color information only. Sorting machines typically detect the color response of products to specific LEDs with various light wavelengths. Unfortunately, a color check is often not enough to detect some very common defects. The shape of a product, instead, reveals many important defects and is highly reliable in detecting external objects mixed with food. Also, shape can be used to take detailed measurements of a product, such as its area, length, width, anisotropy, etc. This paper proposes a complete treatment of the problem of sorting food by its shape. It treats real-world problems such as accuracy, execution time, latency and it provides an overview of a full system used on state-of-the-art measurement machines.

The Social Spider Algorithm (SSA) was introduced based on the information-sharing foraging strategy of spiders to solve the continuous optimization problems. SSA was shown to have better performance than the other state-of-the-art meta-heuristic algorithms in terms of best-achieved fitness values, scalability, reliability, and convergence speed. By preserving all strengths and outstanding performance of SSA, we propose a novel algorithm named Discrete Social Spider Algorithm (DSSA), for solving discrete optimization problems by making some modifications to the calculation of distance function, construction of follow position, the movement method, and the fitness function of the original SSA. DSSA is employed to solve the symmetric and asymmetric traveling salesman problems. To prove the effectiveness of DSSA, TSPLIB benchmarks are used, and the results have been compared to the results obtained by six different optimization methods: discrete bat algorithm (IBA), genetic algorithm (GA), an island-based distributed genetic algorithm (IDGA), evolutionary simulated annealing (ESA), discrete imperialist competitive algorithm (DICA) and a discrete firefly algorithm (DFA). The simulation results demonstrate that DSSA outperforms the other techniques. The experimental results show that our method is better than other evolutionary algorithms for solving the TSP problems. DSSA can also be used for any other discrete optimization problem, such as routing problems.