Conference paper Open Access

Explainable Deep Neural Networks for Multivariate Time Series Predictions

Assaf, Roy; Schumann, Anika


JSON Export

{
  "files": [
    {
      "links": {
        "self": "https://zenodo.org/api/files/c1609803-befd-4cfc-b96e-7b94eba3c96c/0932.pdf"
      }, 
      "checksum": "md5:eb3d3d69332ae51638c10317898a7226", 
      "bucket": "c1609803-befd-4cfc-b96e-7b94eba3c96c", 
      "key": "0932.pdf", 
      "type": "pdf", 
      "size": 1247477
    }
  ], 
  "owners": [
    103348
  ], 
  "doi": "10.24963/ijcai.2019/932", 
  "stats": {
    "version_unique_downloads": 98.0, 
    "unique_views": 126.0, 
    "views": 143.0, 
    "version_views": 143.0, 
    "unique_downloads": 98.0, 
    "version_unique_views": 126.0, 
    "volume": 130985085.0, 
    "version_downloads": 105.0, 
    "downloads": 105.0, 
    "version_volume": 130985085.0
  }, 
  "links": {
    "doi": "https://doi.org/10.24963/ijcai.2019/932", 
    "latest_html": "https://zenodo.org/record/3843273", 
    "bucket": "https://zenodo.org/api/files/c1609803-befd-4cfc-b96e-7b94eba3c96c", 
    "badge": "https://zenodo.org/badge/doi/10.24963/ijcai.2019/932.svg", 
    "html": "https://zenodo.org/record/3843273", 
    "latest": "https://zenodo.org/api/records/3843273"
  }, 
  "created": "2020-05-25T13:10:10.341099+00:00", 
  "updated": "2020-05-25T20:20:24.686233+00:00", 
  "conceptrecid": "3843272", 
  "revision": 2, 
  "id": 3843273, 
  "metadata": {
    "access_right_category": "success", 
    "doi": "10.24963/ijcai.2019/932", 
    "description": "<p>We demonstrate that CNN deep neural networks can not only be used for making predictions based on multivariate time series data, but also for explaining these predictions. This is important for a number of applications where predictions are the basis for decisions and actions. Hence, confidence in the prediction result is crucial. We design a two stage convolutional neural network architecture which uses particular kernel sizes. This allows us to utilise gradient based techniques for generating saliency maps for both the time dimension and the features. These are then used for explaining which features during which time interval are responsible for a given prediction, as well as explaining during which time intervals was the joint contribution of all features most important for that prediction. We demonstrate our approach for predicting the average energy production of photovoltaic power plants and for explaining these predictions.</p>", 
    "license": {
      "id": "CC-BY-4.0"
    }, 
    "title": "Explainable Deep Neural Networks for Multivariate Time Series Predictions", 
    "relations": {
      "version": [
        {
          "count": 1, 
          "index": 0, 
          "parent": {
            "pid_type": "recid", 
            "pid_value": "3843272"
          }, 
          "is_last": true, 
          "last_child": {
            "pid_type": "recid", 
            "pid_value": "3843273"
          }
        }
      ]
    }, 
    "grants": [
      {
        "code": "745625", 
        "links": {
          "self": "https://zenodo.org/api/grants/10.13039/501100000780::745625"
        }, 
        "title": "Reliable OM decision tools and strategies for high LCoE reduction on Offshore wind", 
        "acronym": "ROMEO", 
        "program": "H2020", 
        "funder": {
          "doi": "10.13039/501100000780", 
          "acronyms": [], 
          "name": "European Commission", 
          "links": {
            "self": "https://zenodo.org/api/funders/10.13039/501100000780"
          }
        }
      }
    ], 
    "keywords": [
      "AI: Machine Learning", 
      "AI: Knowledge Representation", 
      "Reasoning Applications: Energy"
    ], 
    "publication_date": "2019-07-01", 
    "creators": [
      {
        "affiliation": "IBM Research Zurich", 
        "name": "Assaf, Roy"
      }, 
      {
        "affiliation": "IBM Research Zurich", 
        "name": "Schumann, Anika"
      }
    ], 
    "access_right": "open", 
    "resource_type": {
      "subtype": "conferencepaper", 
      "type": "publication", 
      "title": "Conference paper"
    }
  }
}
143
105
views
downloads
Views 143
Downloads 105
Data volume 131.0 MB
Unique views 126
Unique downloads 98

Share

Cite as