Saliency methods are used extensively to highlight the importance of input features in model predictions. These methods are mostly used in vision and language tasks, and their applications to time series data is relatively unexplored. In this paper, we set out to extensively compare the performance of various saliency-based interpretability methods across diverse neural architectures, including Recurrent Neural Network, Temporal Convolutional Networks, and Transformers in a new benchmark of synthetic time series data. We propose and report multiple metrics to empirically evaluate the performance of saliency methods for detecting feature importance over time using both precision (i.e., whether identified features contain meaningful signals) and recall (i.e., the number of features with signal identified as important). Through several experiments, we show that (i) in general, network architectures and saliency methods fail to reliably and accurately identify feature importance over time in time series data, (ii) this failure is mainly due to the conflation of time and feature domains, and (iii) the quality of saliency maps can be improved substantially by using our proposed two-step temporal saliency rescaling (TSR) approach that first calculates the importance of each time step before calculating the importance of each feature at a time step.

This paper gives an overview on how to develop a dense and deep neural network for making a time series prediction. First, the history and cornerstones in Artificial Intelligence and Machine Learning will be presented. After a short introduction to the theory of Artificial Intelligence and Machine Learning, the paper will go deeper into the techniques for conducting a time series prediction with different models of neural networks. For this project, Python's development environment Jupyter, extended with the TensorFlow package and deep-learning application Keras is used. The system setup and project framework are explained in more detail before discussing the time series prediction. The main part shows an applied example of time series prediction with weather data. For this work, a deep recurrent neural network with Long Short-Term Memory cells is used to conduct the time series prediction. The results and evaluation of the work show that a weather prediction with deep neural networks can be successful for a short time period. However, there are some drawbacks and limitations with time series prediction, which will be discussed towards the end of the paper.

This method is not developed for time-series, only for tabular data, so it is unclear how they would do this evaluation, neither is it described anywhere in the main draft.

This work introduces a bunch of benchmarks for evaluating time series saliency methods (with respective metrics). The authors do a number of empirical evaluations, draw some conclusions about why certain things don't work, and propose a new saliency method based on that. There are a number of things that I like about this work and that was pointed out by the reviewers as well: there is a definite lack of datasets with groundtruth saliency in them so coming up with such a dataset (and associated metrics) is a worthy contribution by itself (though perhaps not rising up to the bar of acceptance at NeurIPS). In general, everyone agreed that this part of the paper is good. What was more controversial: is the subsequent analysis interesting and novel enough?

Saliency methods are used extensively to highlight the importance of input features in model predictions. These methods are mostly used in vision and language tasks, and their applications to time series data is relatively unexplored. In this paper, we set out to extensively compare the performance of various saliency-based interpretability methods across diverse neural architectures, including Recurrent Neural Network, Temporal Convolutional Networks, and Transformers in a new benchmark of synthetic time series data. We propose and report multiple metrics to empirically evaluate the performance of saliency methods for detecting feature importance over time using both precision (i.e., whether identified features contain meaningful signals) and recall (i.e., the number of features with signal identified as important). Through several experiments, we show that (i) in general, network architectures and saliency methods fail to reliably and accurately identify feature importance over time in time series data, (ii) this failure is mainly due to the conflation of time and feature domains, and (iii) the quality of saliency maps can be improved substantially by using our proposed two-step temporal saliency rescaling (TSR) approach that first calculates the importance of each time step before calculating the importance of each feature at a time step.

As the world increasingly relies on mathematical models for forecasts in different areas, effective communication of uncertainty in time series predictions is important for informed decision making. This study explores how users estimate probabilistic uncertainty in time series predictions under different variants of line charts depicting uncertainty. It examines the role of individual characteristics and the influence of user-reported metrics on uncertainty estimations. By addressing these aspects, this paper aims to enhance the understanding of uncertainty visualization and for improving communication in time series forecast visualizations and the design of prediction data dashboards.As the world increasingly relies on mathematical models for forecasts in different areas, effective communication of uncertainty in time series predictions is important for informed decision making. This study explores how users estimate probabilistic uncertainty in time series predictions under different variants of line charts depicting uncertainty. It examines the role of individual characteristics and the influence of user-reported metrics on uncertainty estimations. By addressing these aspects, this paper aims to enhance the understanding of uncertainty visualization and for improving communication in time series forecast visualizations and the design of prediction data dashboards.

Explaining predictions based on multivariate time series data carries the additional difficulty of handling not only multiple features, but also time dependencies. It matters not only what happened, but also when, and the same feature could have a very different impact on a prediction depending on this time information. Previous work has used perturbation-based saliency methods to tackle this issue, perturbing an input using a trainable mask to discover which features at which times are driving the predictions. However these methods introduce fixed perturbations, inspired from similar methods on static data, while there seems to be little motivation to do so on temporal data. In this work, we aim to explain predictions by learning not only masks, but also associated perturbations. We empirically show that learning these perturbations significantly improves the quality of these explanations on time series data.

How can we explain the predictions of a machine learning model? When the data is structured as a multivariate time series, this question induces additional difficulties such as the necessity for the explanation to embody the time dependency and the large number of inputs. To address these challenges, we propose dynamic masks (Dynamask). This method produces instance-wise importance scores for each feature at each time step by fitting a perturbation mask to the input sequence. In order to incorporate the time dependency of the data, Dynamask studies the effects of dynamic perturbation operators.

How can we explain the predictions of a machine learning model? When the data is structured as a multivariate time series, this question induces additional difficulties such as the necessity for the explanation to embody the time dependency and the large number of inputs. To address these challenges, we propose dynamic masks (Dynamask). This method produces instance-wise importance scores for each feature at each time step by fitting a perturbation mask to the input sequence. In order to incorporate the time dependency of the data, Dynamask studies the effects of dynamic perturbation operators. In order to tackle the large number of inputs, we propose a scheme to make the feature selection parsimonious (to select no more feature than necessary) and legible (a notion that we detail by making a parallel with information theory). With synthetic and real-world data, we demonstrate that the dynamic underpinning of Dynamask, together with its parsimony, offer a neat improvement in the identification of feature importance over time. The modularity of Dynamask makes it ideal as a plug-in to increase the transparency of a wide range of machine learning models in areas such as medicine and finance, where time series are abundant.

Many research papers cover the prediction of financial time series but only a small number of them speak about the application in a real trading strategy. Most of the time the research only gives the performance metrics of the model (accuracy, RMSE, …) but without trying to transform it into a profitable strategy. When we talk about financial time series, we talk about stochastic processes, meaning it deals with a lot of randomness. For this reason, it's unrealistic to expect getting an accuracy similar to the ones obtained in many other applications of Deep Learning. So, don't expect getting a 80% accuracy when predicting the market to go up or down for a given time horizon.