Saudi Cultural Missions Theses & Dissertations
Permanent URI for this communityhttps://drepo.sdl.edu.sa/handle/20.500.14154/10
Browse
3 results
Search Results
Item Restricted Improving Model-fitting for Computational Models of Neural Function and Behaviour(King's College London, 2024) Ramis, Sarah; wise, TobyParameter estimation plays a critical role in computational neuroscience, enabling the design of models that can predict neural mechanisms under various conditions using statistical methods. However, accurate estimation becomes challenging when the likelihood function is intractable due to the model’s complexity and high dimensionality. With the advent of artificial intelligence, various machine learning algorithms have been developed to estimate the posterior distribution of observed data using simulation-based inference approaches. This study aims to elucidate the capabilities of these machine learning models compared to conventional methods to aid in model selection. Rescorla-Wagner simulated data is used to compare the performance of several techniques: Markov Chain Monte Carlo (MCMC) as a conventional method, Automatic Posterior Transformation (APT) using the Masked Auto-regressive Flow model, and APT enhanced with Gated Recurrent Units (APT-GRU) for data embedding, to estimate the parameters alpha and beta. The experiment involved 12 trials with varying parameter pairs and data volumes. The models were evaluated based on computational time, Mean Squared Error (MSE), Mean Absolute Error (MAE), a customized fit metric, Kullback-Leibler (KL) divergence, convergence, and log-likelihood. Results indicated that computational time increased across all models as dataset volume grew, with MCMC requiring significantly more time than the others. APT-GRU achieved lower MSE and MAE scores than the other models, indicating higher accuracy, followed by APT. Both APT-based models demonstrated stable convergence on large datasets, with APT-GRU showing enhanced convergence on smaller datasets. APT models also approached zero in KL divergence scores and positive log-likelihood values, signifying a good model fit. MCMC, while performing better in estimating alpha than beta, showed large negative KL divergence values, suggesting underfitting. Furthermore, MCMC exhibited lower performance stability across trials, with higher estimation variances. In summary, the findings highlight the strengths of APT and APT-GRU in parameter estimation within the Rescorla-Wagner framework. They suggest that the inclusion of GRU for data embedding improves both model robustness and estimation accuracy. These insights pave the way for future research into exploring different machine learning algorithms and model architectures for parameter estimations.12 0Item Restricted Network Intrusion Detection Against Advanced Persistent Threats(Imperial College London, 2024-03-11) Alageel, Almuthanna; Maffeis, SergioThe thesis explores the challenges of detecting Advanced Persistent Threats (APTs) due to their complex nature and low occurrence. The study focuses on network intrusion detection and analyzes 33 APT campaigns spanning the past 22 years. It finds that 81% of APT campaigns use HTTP(S) for evasion techniques, while 45% utilize the DNS protocol for resolution and tunnelling. By analyzing data from 63 APT campaigns over 13 years, we propose HawkEye, a system that achieves an accuracy of 98.53%, a macro average F1-score of 90.38%, and a low false positive rate (FPR) of 0.48% against unseen APT campaigns. In comparison, the baseline achieves lower performance, with accuracy, F1-score, and FPR values of 96.95%, 76.81%, and 0.68%, respectively. The thesis also examines the TTPs used by APTs employing HTTP(S) protocols and introduces EarlyCrow, which achieves a headline macro average F1-score of 93.72%, an accuracy of 98.11%, and an FPR of 0.74% against unseen APTs. On the other hand, the state of the art achieves a 60.29% F1-score with no false positive rates. Additionally, we present NightVision, which extracts information from network traffic using statistical digital signal processing techniques. NightVision achieves an average F1-score of 80.09%, an accuracy rate of 97.71%, and a low FPR of 0.25%. In comparison, the state of the art baseline performs at 67.61% F1-score, 95.82% accuracy, and 1.61% FPR, respectively. We recommend using the proposed tools in conjunction with Host Intrusion Detection Systems (HIDS) to enhance overall security defences against APTs. By combining HawkEye, EarlyCrow, and NightVision, the approach aims to provide a comprehensive and effective defence mechanism.33 0Item Restricted Valuation and Forecasting of Petroleum Assets: A Comprehensive Review and Comparative Analysis of Asset Pricing Models in the Oil Industry(Saudi Digital Library, 2023-11-09) Almayouf, Abdulaziz; Leslie, KirstenThe thesis aims to conduct a comprehensive examination and consolidation of current scholarly works on asset pricing models within the oil industry, employing the PRISMA methodology. Furthermore, this study employs regression analysis and model comparisons utilising Microsoft Excel to assess the efficacy of various asset pricing models in predicting asset prices within the oil industry. The literature review involves conducting a thorough search and analysis of pertinent studies, emphasising the significance of asset pricing models within the oil industry. Arbitrage Pricing Theory (APT), the Capital Asset Pricing Model (CAPM), and the Fama-French Three-Factor Model (FF3) are some theories that have been studied in the literature. The performance evaluation of asset pricing models involves the utilisation of Microsoft Excel for conducting regression analysis and model comparisons. The analysis primarily centres on the assessment of the adequacy of the model fit by utilising statistical measures such as R-squared, adjusted R-squared, and the significance of coefficients. Residual analysis facilitates the evaluation of model performance by enabling the examination of discrepancies between predicted and observed values. The utilisation of the PRISMA methodology in conducting a systematic review allows for the identification and analysis of the strengths, weaknesses, and gaps present within the existing body of literature. This comprehensive evaluation provides practitioners and researchers with the necessary information to make well-informed decisions. The utilisation of regression analysis contributes to the advancement of knowledge by assessing the efficacy of various models, among which the Arbitrage Pricing Theory (APT) model exhibits potential in its ability to forecast asset prices within the oil industry. This thesis underscores the necessity of developing more advanced and customised asset pricing models that incorporate the distinctive attributes of the oil industry. The strengths and weaknesses that have been identified serve as a valuable guide for practitioners in developing risk management strategies. Additionally, the insights obtained from the regression analysis provide valuable information for determining future research directions in asset pricing specifically within the oil industry.9 0