Sindh University Research Journal - SURJ (Science Series)

Contra-Harmonic Mean Derivative-Based Open Newton-Cotes Quadrature Rules

2025-11-05T07:29:55+00:00

A novel family of open Newton-Cotes formulas is created to assess definite integrals. The new family is formed by taking the Contra-Harmonic Mean of the function's first-order derivatives within the interval [a,b]. In comparison to the classical open Newton-Cotes quadrature rules, the proposed derivative-based quadrature rules improve accuracy by two orders of magnitude. These formulas arise from the concept of degree of precision. Furthermore, the theoretical conclusions are validated by calculating the computational order of accuracy for each approach. The computational cost and absolute error drops calculated for three different integrals from the literature illustrate the superiority of the proposed approaches over the classical method.

Hybrid Deep Learning Model for Bitcoin and Ethereum Price Prediction Using Sentiment Analysis

2025-10-08T04:41:35+00:00

Cryptocurrencies have upended the financial industry since they provide decentralized and peer-to-peer transactions. However, due to market volatility and the numerous non-linear relationships between price dynamics and human mood, forecasting Bitcoin values is a difficult task. The deep learning architecture shown in this work combines sentiment confidence scores derived from cryptocurrency-related tweets utilizing Transformer-based natural language processing with historical price indicators. The model incorporates Convolutional Neural Networks (CNN) to detect local time-series patterns and Long Short-Term Memory (LSTM) networks to produce long-term dependencies. We apply this architecture, involving sequence-based preprocessing and normalization, to Bitcoin and Ethereum to ensure robustness. Evaluations in comparison to baseline models Sentiment fusion dramatically increases predicting accuracy, especially during times of market turbulence, according to CNN-LSTM without sentiment, vanilla LSTM, and ARIMA. Our research helps develop scalable, sentiment-aware financial forecasting algorithms that better reflect the behavior of real markets.

Synthesis, characterization of Cu(II)/ Ni(II) complexes metal ions derived from flexible dicarboxylate ligand with 2,2’-bipyridine and their photodegradations applications.

2025-04-29T18:23:52+00:00

Newly developed metal ligand (M-L) coordination complexes were synthesized by various methods for research and commercial uses, which shown significant improvement over time. A newer flexible, bipyridine-based Cu (II) and Ni (II) metal complexes formed from 2,2’-bipyridine-4,4’-dicarboxylic acid, using the metal-coupling reaction. The structure of the new M-L complexes was analyzed using various spectroscopic techniques, including FT-IR, UV/Vis, SEM, and EDS. The newly synthesized compound 2,2’-bipyridine-4,4’-dicarboxylic acid[(BPyCOOH)] showed weak to medium-intensity bands in the 1730 cm?¹ range, attributed to aromatic carboxylate stretching vibrations. These metal complexes were employed as photo-catalysts for the photo degradation of methylene blue in an aqueous solution. The photocatalytic experiments revealed that the Cu(II) complex exhibited the best photocatalytic performance than Ni(II) complex, degrading 88% of methylene blue (MB) at 26°C. The photo-degradation activity was observed under UV light and in darkness. The kinetics of the photo degradation of methylene blue were investigated with optimized parameters in the presence of nan catalysts. The rate constants (k) for the metal ions [(BPyCOO)Cu] and [(BPyCOO)Ni] were measured to be 0.0481 and 0.0212, respectively. These results indicate a pseudo-first-order reaction with a higher rate constant for [(BPyCOO)Cu] compared to [(BPyCOO)Ni].

Soybean Crop Growth and Yield Prediction using Deep Learning Model

2025-07-18T07:39:50+00:00

Agriculture is a fundamental pillar of food security and economic development, with soybean recognized globally for its high nutritional value and versatile applications in both food and non-food industries. Accurate soybean yield forecasting is critical for addressing key agricultural challenges, including efficient resource utilization, sustainable production, and climate change adaptation. Recent advancements in deep learning and remote sensing technologies, incorporating multi-source data such as UAV imagery, satellite data, and field measurements, have significantly enhanced yield prediction accuracy. This research proposes advanced deep learning models, utilizing spatial-temporal frameworks like CNN and LSTM, for precise, fast, and non-intrusive soybean yield assessment. Challenges such as dataset availability, computational demands, and regional variability in soil and climate are addressed to improve model flexibility and scalability. Furthermore, integrating hyperspectral, thermal, and RGB imagery enhances stress detection. This study contributes towards developing scalable, decision-support tools for sustainable and precision agriculture.

Enhancing Urban Sound Classification with CNN-Transformer Hybrid Model and Spectrogram Augmentation

2025-07-21T01:52:00+00:00

Urban Sound Classification (USC) is a crucial component of audio recognition systems, with applications in smart cities, surveillance, and multimedia. Despite significant advances, the classification of environmental sounds remains a challenge due to the complex nature of urban audio signals, characterized by high intra-class variability and overlapping sound events. In this paper, we propose a novel hybrid model that integrates the strengths of Convolutional Neural Networks (CNNs) and Transformer architectures to improve the identification accuracy of urban sounds. The CNN component effectively extracts local spectral features from Mel spectrograms, while the Transformer captures global temporal dependencies through self-attention mechanisms. Additionally, we incorporate advanced spectrogram augmentation techniques, such as time masking, frequency masking, and time warping, to further enhance the model’s robustness and generalization capabilities. Experimental results on the UrbanSound8K dataset demonstrate that the proposed CNN- Transformer hybrid model outperforms traditional CNN and Long Short-Term Memory (LSTM)-based approaches, achieving a classification accuracy of 93.36%. These results highlight the effectiveness of combining CNNs with transformers and data augmentation strategies for robust urban sound classification.

Evaluating the Efficacy of Simulated User Models in Interactive Information Retrieval: A User-Based Approach

2025-06-24T06:39:31+00:00

Abstract

Simulated user models are increasingly employed to evaluate IIR systems due to their scalability and consistency. However, the extent to which these models realistically replicate human behavior across diverse search tasks remains underexplored. This study investigates the behavioral fidelity of simulated users, specifically rule-based and LLM-driven agents, by comparing them to real users across factual, exploration, and comparative search tasks. Using a controlled experimental framework and the Search data set, analyze the 32 real-user sessions and 32 matched simulations based on retrieval performance (MAP, nDCG), behavioral patterns such as query reformulations, session time, and satisfaction measures. The study results show that simulated users closely approximate real-user performance in factual tasks; they significantly underperform in exploratory and comparative contexts, particularly in query reformulation frequency and satisfaction alignment. Simulated satisfaction scores, estimated through relevance proxies, diverged from real user ratings (3.4 vs. 4.1 average), highlighting cognitive and affective realism gaps. These findings suggest that current simulation models lack real users' adaptability and strategic diversity, especially in open-ended tasks. The study contributes empirical evidence of simulation limitations and guides for improving user model fidelity, emphasizing the need for hybrid evaluation frameworks that combine real-user insight with scalable simulation.