3. Electronic Theses and Dissertations (ETDs) - All submissions
Permanent URI for this communityhttps://wiredspace.wits.ac.za/handle/10539/45
Browse
2 results
Search Results
Item Investigating explainablity methods in recurrent neural network architectures for financial time-series data(2022) Freeborough, WarrenStatistical methods were traditionally used for time series forecasting. However, new hybrid methods demonstrate competitive accuracy, leading to increased machine learning-based methodologies in the financial sector. However, very little development has been seen in explainable AI (XAI) for financial time series prediction, with a growing mandate for explainable systems. This study aims to determine if the existing XAI methodology is transferable to the context of financial time series prediction. Four popular methods, namely: ablation, permutation, added noise, integrated gradients, were applied to an RNN, LSTM, and a GRU network trained S&P 500 stocks data to determine the importance of features, individual data points and specific cells in each architecture. The explainability analysis reveals that GRU displayed the most significant ability to retain long-term information, while the LSTM disregarded most of the given input and instead showed the most notable granularity to the considered inputs. Lastly, the RNN displayed features indicative of no long-term memory retention. The applied XAI methods produced complementary results, reinforcing paradigms on significant differences in how different architectures predict. The results show that these methods are transferable in the financial forecasting sector, but a more sophisticated hybrid prediction system requires further confirmationItem The transcription factor interacting network of tolerant TME3 and susceptible T200 cassava landraces infected with SACMV(2019) Freeborough, WarrenCassava, Manihot esculenta Crantz, is categorized as a food security crop, producing large starchy tubers that are gaining interest from both international and local agro-processing industries for products such as bioethanol, textiles, and food additives. However, cassava is currently under threat from a group of begomoviruses that cause cassava mosaic disease (CMD) in all countries in sub-Saharan Africa where cassava is cultivated. CMD can result in up to 100% crop loss. South African cassava mosaic virus (SACMV) is particularly a threat to the growing cassava industry in southern Africa. Despite extensive breeding programs over the past 70 years to develop CMD-resistant farmer-preferred cassava landraces, total resistance has not been achieved. Furthermore, the high mutational rates of begomoviruses, and mixed infections in the field, have exacerbated the problem. TME3 is a West African landrace that displays tolerance to begomoviruses, including SACMV. Infection of TME3 by SACMV leads to recovery, hallmarked by low virus loads and milder symptoms compared to a susceptible southern African landrace T200. The molecular processes that govern tolerance in crops, including cassava, are not well understood. However, systemic immune responses, which are controlled by hormoneresponsive transcription factors (TFs), are required by the plant to successfully combat an invading pathogen. Two different branches of systemic immunity have been described, namely systemic acquired resistance (SAR), facilitated by salicylic acid (SA) signalling, and induced systemic resistance (ISR), which is induced through jasmonic acid (JA) and ethylene (ET) signalling in the presence of beneficial rhizobacteria. In 2014, Allie et al. compared global transcriptomic responses occurring in TME3 and the T200 during early 12 days’ post inoculation (dpi), middle (32 dpi) and late (67 dpi) stages of SACMV infection. In order to give greater context to transcriptomic data, which is inheritably large and complex, network analysis may be implemented. By placing the differentially expressed (DE) gene homologs/orthologs identified from the cassava transcriptome datasets into protein-protein networks, functions of SACMV-responsive genes, interacting partners, and potential hubs, can be derived. Cassava gene functions are based on the model crop Arabidopsis thaliana, as despite the sequencing of the cassava genome, the annotations are incomplete. The aim of this study was to identify potential candidate TFs, and their associated hormones and other network partners, that confer either tolerance (TME3) or susceptibility (T200) to SACMV.