mTSeer: Interactive Visual Exploration of Models on Multivariate Time-series Forecast

Yuan, Jun; Yifang, Wang; Silva, Claudio; Bertini, Enrico

doi:10.1145/3411764.3445083

Cited by 15 publications

(9 citation statements)

References 43 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…(b) Xu et al [41] show the additive attributions of SHAP [14] as bar charts on the forecast target and further arrows to compare multiple models attributions. [19] and Xu et al [41] visualizations for time series attributions focusing on either a pipe enhancement for line plots to visualize relevance of SHAP or multiple models using the additive attribution SHAP technique for an improved comparison between various models. Mujkanovic et al [19] extend the SHAP technique for time series and visualize the corresponding relevance as pipes around the line plot with a color scale.…”

Section: Line Plots With Attribution Extensionsmentioning

confidence: 99%

“…Mujkanovic et al [19] extend the SHAP technique for time series and visualize the corresponding relevance as pipes around the line plot with a color scale. Xu et al [41] incorporate the additive properties of SHAP to compare the performances of time series forecasting models by enabling explanations of single models and a direct comparison with arrows showing the differences in the attributions. challenges can get mitigated by easy-to-understand visualizations on line plots with aggregations such as Mujkanovic et al [19].…”

Section: Line Plots With Attribution Extensionsmentioning

confidence: 99%

“…Another approach by Xu et al [41] incorporates the additive properties of SHAP [14] into their visualization to exchange the heatmap with bar charts on top and beneath a line representing the time series data seen in Fig. 3b.…”

Section: Future Opportunitiesmentioning

confidence: 99%

“…Figure 3: Mujkanovic et al[19] and Xu et al[41] visualizations for time series attributions focusing on either a pipe enhancement for line plots to visualize relevance of SHAP or multiple models using the additive attribution SHAP technique for an improved comparison between various models. Mujkanovic et al[19] extend the SHAP technique for time series and visualize the corresponding relevance as pipes around the line plot with a color scale.…”

mentioning

confidence: 99%

See 3 more Smart Citations

Time Series Model Attribution Visualizations as Explanations

Schlegel¹,

Keim²

2021

Preprint

View full text Add to dashboard Cite

Attributions are a common local explanation technique for deep learning models on single samples as they are easily extractable and demonstrate the relevance of input values. In many cases, heatmaps visualize such attributions for samples, for instance, on images. However, heatmaps are not always the ideal visualization to explain certain model decisions for other data types. In this review, we focus on attribution visualizations for time series. We collect attribution heatmap visualizations and some alternatives, discuss the advantages as well as disadvantages and give a short position towards future opportunities for attributions and explanations for time series.

show abstract

Section: Line Plots With Attribution Extensionsmentioning

confidence: 99%

Section: Line Plots With Attribution Extensionsmentioning

confidence: 99%

Section: Future Opportunitiesmentioning

confidence: 99%

mentioning

confidence: 99%

See 2 more Smart Citations

Time Series Model Attribution Visualizations as Explanations

Schlegel¹,

Keim²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Some previous studies have provided reasonable and explainable predictions for sales volume by comparing different machine learning (ML) and deep learning (DL) models to visually dig into the reasons behind each rise in sales [31,43,47,54]. However, to the best of our knowledge, most of these works have focused on the statistical characteristics of time series data rather than incorporating the promotion strategies that most e-commerce retailers must consider in order to gain more attention and expect higher profits.…”

mentioning

confidence: 99%

PromotionLens: Inspecting Promotion Strategies of Online E-commerce via Visual Analytics

Zhang¹,

Wang²,

Zhao³

et al. 2022

Preprint

View full text Add to dashboard Cite

Fig. 1. Case I: (1) Select a product within the red circle among all 361°sneakers in the Product Overview. (2) Browse the product description of the selected product at the buttom of the scatter chart. (3) View the annual data of the selected product in the Promotion Overview. (4) Adjust the time period selector in the Sales Prediction View. (5) Look through the dotted lines and find non-overlapping zones. (6) Change the promotion strategy in the Strategy View -exchange the value discount that exists with the same discount rate and percentage discount for the time period. (7) Discover that sales have decreased. (8) Conclude that the value discounts may be appropriate for the selected product. (9) Compare similar products in the glyphs, word clouds and box plots in the Competitive Analysis View, and select another XTEP product. (10) Adjust the time selector to fix a time period for the newly selected product. (11) Replace the existing percentage discount in the Strategy View with the same discount rate and value discount for the time period. (12) Find that sales have decreased. ( 13) Conclude that the percentage discount may be appropriate for the selected product.

show abstract