Facebook and Twitter, as two known social media become popular sources of big data that give the right to people to share and express their feedback about products, services, politicians, events, and every aspect of life in the form of short texts. The classification of sentiments could be automated through machine learning and enhanced using appropriate feature extraction methods. In this work, we collected the most recent tweets about (Biden, Benzema, Apple, and NASA) using Twitter-API and assigned sentiment scores using a rule-based lexicon approach; after pre-processing stage, each dataset is divided into 80% as a training set, and rest 20% as testing set. After that, the Distributed bag of words, Distributed memory mean, Distributed Memory Concatenation, and Term Frequency-Inverse Document Frequency models are used for feature extraction from pre-processed tweets. Depending on the Shark smell optimizer algorithm, the SVM technique was used to classify the extracted features. The SSO was used to tune and select the best value for SVM parameters to optimize the overall model performance. The results display that these optimizers have an essential impact on increasing the model accuracy. After optimization, the model accuracy reached 92.12%, while the highest accuracy without optimization was 88.69% for various feature extraction methods.
1 November 2025 in Passer Journal of Basic and Applied Sciences
This research presents a detailed scientific review, that focuses on the sentiment analysis within Kurdish social media, which includes both traditional Machine Learning (ML) and advanced Deep Learning (DL) algorithms. The literature evaluated reveals significant advances in resource development, including the establishment of annotated corpora and benchmark datasets. While overcoming inherent challenges associated with processing a low-resource language characterized by dialectal variation, morphological complexity, and cultural subtleties. Traditional methodologies, such as Naive Bayes and Support Vector Machine (SVM), have demonstrated efficacy in contexts where data availability is limited. While recent advances in Transfer Learning (TL) and transformer-based architectures, exemplified by Kurdish Bidirectional Encoder Representations from Transformers (KuBERT), have shown progress. The creation of annotated datasets, like Kurdish Sentiment (KurdiSent) and Kurdish News Dataset Headlines (KNDH), and the modification of DL techniques to improve performance on several sentiment analysis tasks have advanced significantly. Furthermore, the integration of domain-specific preprocessing strategies spanning customized stemming to the examination of nonverbal indicators such as emoji emphasizes the need for linguistic and cultural customization in model architecture. This review not only validates contemporary developments in Kurdish sentiment analysis but also pinpoints crucial deficiencies, proposing future trajectories for the augmentation of datasets, formulating language-specific Natural Language Processing (NLP) models, multimodal analysis, and standardized assessment frameworks. The insights outlined here aspire to inform and direct future research initiatives aimed at progressing Kurdish NLP and enriching understanding of sentiment dynamics in the context of digital communication.
27 November 2022 in Kurdistan Journal of Applied Research
There are many potential sources of data uncertainty, such as imperfect measurement or sampling, intrusive environmental monitoring, unreliable sensor networks, and inaccurate medical diagnoses. To avoid unintended results, data mining from new applications like sensors and location-based services needs to be done with care. When attempting to classify data with a high degree of uncertainty, many researchers have turned to heuristic approaches and machine learning (ML) methods. We propose an entirely new ML method in this paper by fusing the Radial Basis Function (RBF) network based on ant colony optimization (ACO). After introducing a large amount of uncertainty into a dataset, we normalize the data and finish training on clean data. The ant colony optimization algorithm is then used to train a recurrent neural network. Finally, we evaluate our proposed method against some of the most popular ML methods, including a k-nearest neighbor, support vector machine, random forest, decision tree, logistic regression, and extreme gradient boosting (Xgboost). Error metrics show that our model significantly outperforms the gold standard and other popular ML methods. Using industry-standard performance metrics, the results of our experiments show that our proposed method does a better job of classifying uncertain data than other methods
2 October 2022 in Kurdistan Journal of Applied Research
The stegangraphic system supply premium secrecy and ability of conserving the mystery information from gaining stalked or cracked. The suggested method consists of three phases which are edge detection, embedding and extraction. This paper concentrated on three basic and significant parts which are payload, quality, and security also introduces a new steganography method by using edge detection method and shark smell optimization to effectively hide data with in images. Firstly, to promote the hiding ability and to realize altitude standard of secrecy the mystery message is separated into four parts and the cover image is masked and also divided into four sections, then the edge detection algorithm and shark smell optimization is performed on each section respectively. Edge prospectors were utilized to produce edge pixels in every section to hide mystery message and attain the best payload. To increase security, the shark smell optimization is used to select the best pixels among edge pixels based on its nature in motion, then reflect these pixels above original carrier media. Finally the mystery message bits are hidden in the selected edge pixels by using lest significant bit technique. The experimental outcomes appreciated utilizing several image fitness appreciation fashion, it displays best hiding ability, achieve higher image quality with least standard of deformation and provide altitude standard of secrecy, also the results shows that the suggested method exceeds previous approaches in idioms of the PSNSR, MSE also demonstrate that the mystery information cannot be retrieved of the stego image without realizing the algorithms and the values of parameters that are used in hidden process