Statistical methods were traditionally primarily 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, and integrated gradients, were applied to a recurrent neural network (RNN), long short-term memory (LSTM), and a gated recurrent unit (GRU) network trained on S&P 500 stocks data to determine the importance of features, individual data points, and specific cells in each architecture. The explainability analysis revealed 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 confirmation.

Hybrid methods have been shown to outperform pure statistical and pure deep learning methods at forecasting tasks and quantifying the associated uncertainty with those forecasts (prediction intervals). One example is Exponential Smoothing Recurrent Neural Network (ES-RNN), a hybrid between a statistical forecasting model and a recurrent neural network variant. ES-RNN achieves a 9.4% improvement in absolute error in the Makridakis-4 Forecasting Competition. This improvement and similar outperformance from other hybrid models have primarily been demonstrated only on univariate datasets. Difficulties with applying hybrid forecast methods to multivariate data include (i) the high computational cost involved in hyperparameter tuning for models that are not parsimonious, (ii) challenges associated with auto-correlation inherent in the data, as well as (iii) complex dependency (cross-correlation) between the covariates that may be hard to capture. This paper presents Multivariate Exponential Smoothing Long Short Term Memory (MES-LSTM), a generalized multivariate extension to ES-RNN, that overcomes these challenges. MES-LSTM utilizes a vectorized implementation. We test MES-LSTM on several aggregated coronavirus disease of 2019 (COVID-19) morbidity datasets and find our hybrid approach shows consistent, significant improvement over pure statistical and deep learning methods at forecast accuracy and prediction interval construction.

The techniques of hybridisation and ensemble learning are popular model fusion techniques for improving the predictive power of forecasting methods. With limited research that instigates combining these two promising approaches, this paper focuses on the utility of the Exponential Smoothing-Recurrent Neural Network (ES-RNN) in the pool of base learners for different ensembles. We compare against some state-of-the-art ensembling techniques and arithmetic model averaging as a benchmark. We experiment with the M4 forecasting dataset of 100,000 time-series, and the results show that the Feature-Based FORecast Model Averaging (FFORMA), on average, is the best technique for late data fusion with the ES-RNN. However, considering the M4’s Daily subset of data, stacking was the only successful ensemble at dealing with the case where all base learner performances were similar. Our experimental results indicate that we attain state-of-the-art forecasting results compared to Neural Basis Expansion Analysis (N-BEATS) as a benchmark. We conclude that model averaging is a more robust ensembling technique than model selection and stacking strategies. Further, the results show that gradient boosting is superior for implementing ensemble learning strategies.

Graph neural networks (GNNs) build on the success of deep learning models by extending them for use in graph spaces. Transfer learning has proven extremely successful for traditional deep learning problems, resulting in faster training and improved performance. Despite the increasing interest in GNNs and their use cases, there is little research on their transferability. This research demonstrates that transfer learning is effective with GNNs, and describes how source tasks and the choice of GNN impact the ability to learn generalisable knowledge. We perform experiments using real-world and synthetic data within the contexts of node classification and graph classification. To this end, we also provide a general methodology for transfer learning experimentation and present a novel algorithm for generating synthetic graph classification tasks. We compare the performance of GCN, GraphSAGE and GIN across both synthetic and real-world datasets. Our results demonstrate empirically that GNNs with inductive operations yield statistically significantly improved transfer. Further, we show that similarity in community structure between source and target tasks support statistically significant improvements in transfer over and above the use of only the node attributes.

Galaxy morphology characterisation is an important area of study, as the type and formation of galaxies offer insights into the origin and evolution of the universe. Owing to the increased availability of images of galaxies, scientists have turned to crowd-sourcing to automate the process of instance labelling. However, research has shown that using crowd-sourced labels for galaxy classification comes with many pitfalls. An alternative approach to galaxy classification is metric learning. Metric learning allows for improved representations for classification, anomaly detection, information retrieval, clustering and dimensionality reduction. Understanding the implications of this approach regarding crowd-sourced labels is of paramount importance if scientists intend to continue using them. This paper compares metric learning and classification models trained or fine-tuned on both the crowd-sourced Galaxy Zoo 2 (GZ2-H) dataset and expertly labelled EFIGI catalogue. The study uses the Revised Shapley-Ames (RSA) catalogue of bright galaxies, also labelled by experts, as an unseen test set. The RSA catalogue allows for an accurate comparison of the performance of the models at predicting the Hubble types of galaxies. The classification accuracy for the crowd-sourced and expert models indicated that the models are comparable on the surface. However, using alternative metrics, the results show that the models trained on the expert dataset outperformed the model trained on the crowd-sourced data in terms of actual vs predicted labels. Further, the results show that fine-tuning a model pre-trained on crowd-sourced data can outperform the state-of-the-art in galaxy characterisation. The models trained to predict the Hubble types of galaxies are better when fine-tuned using the Proxy-NCA and Normalised-Softmax loss functions than with other pairwise losses. The Normalised-Softmax loss yielded the best overall 9-class models with accuracies at 30.88% (GZ2-H) and 30.05% (EFIGI) and MAP values of 0.3483 (GZ2-H) and 0.3889. The Proxy-NCA loss produced the second-best overall 9-class models with accuracies at 30.33% (GZ2-H) and 20.03% (EFIGI) and MAP values of 0.3577 (GZ2-H) and 0.3917 (EFIGI). Finally, the paper highlights the need for caution when utilising crowd-sourced labels; however, it argues that transfer learning from crowd-sourced labelled data to expert-labelled data can still lead to significant improvements.