Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network
Testing of evidence in criminal cases can be limited by temporal or financial constraints or by the fact that certain tests may be mutually exclusive, so choosing the tests that will have maximal impact on the final result is essential. In this paper, we assume that a main hypothesis, evidence for i...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2018-11-01
|
| Series: | Entropy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1099-4300/20/11/856 |
| _version_ | 1798040460612075520 |
|---|---|
| author | Leila Schneps Richard Overill David Lagnado |
| author_facet | Leila Schneps Richard Overill David Lagnado |
| author_sort | Leila Schneps |
| collection | DOAJ |
| description | Testing of evidence in criminal cases can be limited by temporal or financial constraints or by the fact that certain tests may be mutually exclusive, so choosing the tests that will have maximal impact on the final result is essential. In this paper, we assume that a main hypothesis, evidence for it and possible tests for existence of this evidence are represented in the form of a Bayesian network, and use three different methods to measure the impact of a test on the main hypothesis. We illustrate the methods by applying them to an actual digital crime case provided by the Hong Kong police. We conclude that the Kullback⁻Leibler divergence is the optimal method for selecting the tests with the highest impact. |
| first_indexed | 2024-04-11T22:07:54Z |
| format | Article |
| id | doaj.art-554f80b393e94cacb96ac134b9342a1e |
| institution | Directory Open Access Journal |
| issn | 1099-4300 |
| language | English |
| last_indexed | 2024-04-11T22:07:54Z |
| publishDate | 2018-11-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Entropy |
| spelling | doaj.art-554f80b393e94cacb96ac134b9342a1e2022-12-22T04:00:38ZengMDPI AGEntropy1099-43002018-11-01201185610.3390/e20110856e20110856Ranking the Impact of Different Tests on a Hypothesis in a Bayesian NetworkLeila Schneps0Richard Overill1David Lagnado2Institut de Mathématiques de Jussieu, Paris 75013, FranceDepartment of Informatics, King’s College London, London WC2R 2LS, UKDepartment of Experimental Psychology, University College London, London WC1H 0AP, UKTesting of evidence in criminal cases can be limited by temporal or financial constraints or by the fact that certain tests may be mutually exclusive, so choosing the tests that will have maximal impact on the final result is essential. In this paper, we assume that a main hypothesis, evidence for it and possible tests for existence of this evidence are represented in the form of a Bayesian network, and use three different methods to measure the impact of a test on the main hypothesis. We illustrate the methods by applying them to an actual digital crime case provided by the Hong Kong police. We conclude that the Kullback⁻Leibler divergence is the optimal method for selecting the tests with the highest impact.https://www.mdpi.com/1099-4300/20/11/856Bayesian networksimpact measuresKullback–Leibler divergencetornado method |
| spellingShingle | Leila Schneps Richard Overill David Lagnado Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network Entropy Bayesian networks impact measures Kullback–Leibler divergence tornado method |
| title | Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network |
| title_full | Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network |
| title_fullStr | Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network |
| title_full_unstemmed | Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network |
| title_short | Ranking the Impact of Different Tests on a Hypothesis in a Bayesian Network |
| title_sort | ranking the impact of different tests on a hypothesis in a bayesian network |
| topic | Bayesian networks impact measures Kullback–Leibler divergence tornado method |
| url | https://www.mdpi.com/1099-4300/20/11/856 |
| work_keys_str_mv | AT leilaschneps rankingtheimpactofdifferenttestsonahypothesisinabayesiannetwork AT richardoverill rankingtheimpactofdifferenttestsonahypothesisinabayesiannetwork AT davidlagnado rankingtheimpactofdifferenttestsonahypothesisinabayesiannetwork |