Loading...
Please wait, while we are loading the content...
Similar Documents
Arctic ice, george clooney, lipstick on a pig, and insomniac fruit flies: combining kd and m&s for predictive analysis
| Content Provider | ACM Digital Library |
|---|---|
| Author | Colbaugh, Richard |
| Abstract | Both knowledge discovery (KD) and modeling and simulation (M&S) have made profound contributions to human understanding, and given their complementary perspectives and shared emphasis on "real-world" phenomena it is natural to suspect that they may be even more powerful when applied in combination. However, the KD and M&S communities have operated and evolved essentially independently, so that this possibility remains largely unexplored. This talk will illustrate, through a series of case studies taken from the predictive analysis domain, some of the substantial benefits of combining these two approaches. I will begin by briefly reviewing two climate dynamics studies which serve as exemplars of two "standard" ways KD and M&S can work together: 1.) KD can uncover regularities and patterns in data which can be incorporated into computational models, 2.) the outputs of (large) M&S runs can be analyzed using KD techniques, enabling rigorous assessments as well as new and unexpected insights. I will then propose two additional ways to integrate KD and M&S. The first is motivated by a key challenge encountered in social dynamics prediction problems: identifying system features which possess predictive power. Applying M&S to this task enables the discovery of predictive features which would be difficult or impossible to uncover using standard feature selection tools; these features are then employed for KD based predictive analysis. This approach is illustrated through a case study involving predictions of the collective human dynamics associated with movie attendance, political campaigns, epidemics, and protests. This integration of M&S and KD reveals that system attributes commonly believed to be predictive for social dynamics are frequently not, and that subtle features of social network dynamics often have considerable predictive power. The second approach to combining KD and M&S is to use M&S tools to develop mathematical models for the system of interest and then apply KD methods to the models to form predictions without running simulations. This methodology is illustrated through a case study involving the detection of vulnerabilities, and prediction of their consequences, in complex biological and technological networks (e.g., circadian rhythm gene networks, electric power grids). An important benefit of this approach is the possibility to obtain provably-correct characterizations of network dynamics, as this permits rigorous, comprehensive predictions of network behavior. |
| Starting Page | 1 |
| Ending Page | 1 |
| Page Count | 1 |
| File Format | |
| ISBN | 9781450308366 |
| DOI | 10.1145/2023568.2023570 |
| Language | English |
| Publisher | Association for Computing Machinery (ACM) |
| Publisher Date | 2011-08-21 |
| Publisher Place | New York |
| Access Restriction | Subscribed |
| Subject Keyword | Modeling and simulation Knowledge discovery Human understanding |
| Content Type | Text |
| Resource Type | Article |
