## Uncertain Reasoning and Forecasting (1995)

Venue: | International Journal of Forecasting |

Citations: | 21 - 3 self |

### BibTeX

@ARTICLE{Dagum95uncertainreasoning,

author = {Paul Dagum and Adam Galper and Eric Horvitz and Adam Seiver},

title = {Uncertain Reasoning and Forecasting},

journal = {International Journal of Forecasting},

year = {1995},

volume = {11},

pages = {73--87}

}

### OpenURL

### Abstract

We develop a probability forecasting model through a synthesis of Bayesian beliefnetwork models and classical time-series analysis. By casting Bayesian time-series analyses as temporal belief-network problems, weintroduce dependency models that capture richer and more realistic models of dynamic dependencies. With richer models and associated computational methods, we can movebeyond the rigid classical assumptions of linearityin the relationships among variables and of normality of their probability distributions.

### Citations

7682 |
Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference
- Pearl
- 1988
(Show Context)
Citation Context ...rtain knowledge. A Bayesian belief network is an expressive knowledge representation for uncertain reasoning that employs a graphical structure to capture explicit dependencies among domain variables =-=[35, 36]-=-. Belief networks are the basis of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for ... |

1374 |
Local computations with probabilities on graphical structures and their applications to expert systems
- Lauritzen, Spiegelhalter
- 1988
(Show Context)
Citation Context ...ng domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks =-=[8, 12, 16, 27, 34]-=- and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance... |

1165 | A Bayesian method for the induction of probabilistic networks from data
- Cooper, Herskovits
- 1992
(Show Context)
Citation Context ...of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks =-=[11, 37, 38, 43]-=- have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance of graphical structures and the fundamental notion of co... |

619 |
The computational complexity of probabilistic inference using Bayesian belief networks
- Cooper
- 1990
(Show Context)
Citation Context ...alue e. Probabilistic inference in large multiply-connected belief networks is di cult. Complexity analyses show that both exact and approximate algorithms pose intractable problems in the worst case =-=[10, 17]-=-. Nevertheless, for many problems, inference approximation procedures provide useful estimates of posterior probabilities in acceptable computation times. 3.2 Additive Belief-Network Models There are ... |

497 |
A Model for Reasoning about Persistence and Causation
- Dean, Kanazawa
- 1989
(Show Context)
Citation Context ...ter, a state-space forecasting model. Since then, few techniques have been developed to address the forecasting problem directly. Several approaches for probabilistic reasoning about change over time =-=[18, 30]-=- and for temporal reasoning with belief networks and in uence diagrams [4, 44, 39] have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices ... |

479 |
Graphical Models in applied multivariate statistics
- Whittaker
- 1990
(Show Context)
Citation Context ...s. More generally, applied statisticians have recognized the importance of graphical structures and the fundamental notion of conditional independence in the multivariate analysis of categorical data =-=[45]-=-. Forecasting has been addressed only intermittently over the last decade by the uncertain reasoning community [19]. Early work on continuous-variable in uence diagrams by Kenley [31] yielded an in ue... |

325 |
Planning and Control
- Dean, Wellman
- 1991
(Show Context)
Citation Context ... notion of conditional independence in the multivariate analysis of categorical data [45]. Forecasting has been addressed only intermittently over the last decade by the uncertain reasoning community =-=[19]-=-. Early work on continuous-variable in uence diagrams by Kenley [31] yielded an in uence-diagram representation of the Kalman lter, a state-space forecasting model. Since then, few techniques have bee... |

266 | Approximating probabilistic inference in Bayesian belief networks is NP-hard
- Dagum, Luby
- 1993
(Show Context)
Citation Context ...ng domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks =-=[8, 12, 16, 27, 34]-=- and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance... |

259 | Bayesian networks without tears
- Charniak
- 1991
(Show Context)
Citation Context ...ain reasoning that employs a graphical structure to capture explicit dependencies among domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications =-=[2, 3, 7, 26, 28, 29, 42]-=-. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network app... |

252 |
1989]: â€˜The ALARM Monitoring System: A Case Study with Two Probabilistic Inference Techniques for Belief Networks
- Beinlich, Suermondt, et al.
- 1989
(Show Context)
Citation Context ...ain reasoning that employs a graphical structure to capture explicit dependencies among domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications =-=[2, 3, 7, 26, 28, 29, 42]-=-. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network app... |

228 |
Combining probability distributions: A critique and an annotated bibliography
- Genest, Zidek
- 1986
(Show Context)
Citation Context ...me Y and the sets of causes X1; :::; Xk, represents a model for combining probabilities. The theory of combining probability distributions has been extensively studied by the Bayesian community (eg., =-=[22]-=-). These combination rules range from linear combinations to multiplicative combinations, depending on the mutual dependence of the various alternative forecasts. This mutual dependence re ects the ca... |

218 | A theory of inferred causation
- Pearl, Verma
- 1991
(Show Context)
Citation Context ...of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks =-=[11, 37, 38, 43]-=- have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance of graphical structures and the fundamental notion of co... |

155 |
Probabilistic Reasoning in Expert Systems
- Neapolitan
- 1990
(Show Context)
Citation Context ...rtain knowledge. A Bayesian belief network is an expressive knowledge representation for uncertain reasoning that employs a graphical structure to capture explicit dependencies among domain variables =-=[35, 36]-=-. Belief networks are the basis of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for ... |

91 |
MUNIN--a causal probabilistic network for interpretation of electromyographic findings
- Andreassen, Woldbye, et al.
- 1987
(Show Context)
Citation Context ...ain reasoning that employs a graphical structure to capture explicit dependencies among domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications =-=[2, 3, 7, 26, 28, 29, 42]-=-. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network app... |

85 | Toward normative expert systems: Part I. The Pathfinder project
- Heckerman, Horvitz, et al.
- 1992
(Show Context)
Citation Context |

83 |
A Computational Model for Causal and Diagnostic Reasoning in Inference Engines
- Kim, Pearl
- 1983
(Show Context)
Citation Context ...recast models; real-world applications of belief-network models also encounter these problems. These limitations prompted the development of models of causal independence and disjunctive interactions =-=[23, 32, 36]-=- that reduce substantially the storage requirements of conditional probability tables. These models, however, do not address the problem of intractable inference. Elsewhere, we develop additive belief... |

81 |
Decision Analysis and Expert Systems
- Henrion, Breese, et al.
- 1991
(Show Context)
Citation Context |

73 |
Linear smoothers and additive models (with discussion), Annals of Statistics 17
- Buja, Hastie, et al.
- 1989
(Show Context)
Citation Context ... approach that reduces the speci cation of an extremely large contingency table into the speci cation of a few very small tables [13]. These models belong to the more general class of additive models =-=[5, 24, 25]-=- that approximate multivariate functions by sums of univariate functions. Fortuitously, additive belief-network models also overcome the intractability of probabilistic inference, and thus they provid... |

67 |
Dynamic network models for forecasting, in
- Dagum, Galper, et al.
(Show Context)
Citation Context ...asting with belief networks include forecasting crude-oil prices [1]. We have developed a probability forecasting model through a synthesis of belief-network models and classical time-series analysis =-=[14, 15]-=-. The dynamic network model (DNM) is based 2son the integration of Bayesian time-series analysis with generalizations of belief-network representation and inference techniques [13]. DNMs inherit the e... |

44 |
Generalized additive models (with discussion
- TRIPATHI, HASTIE, et al.
- 1986
(Show Context)
Citation Context ... approach that reduces the speci cation of an extremely large contingency table into the speci cation of a few very small tables [13]. These models belong to the more general class of additive models =-=[5, 24, 25]-=- that approximate multivariate functions by sums of univariate functions. Fortuitously, additive belief-network models also overcome the intractability of probabilistic inference, and thus they provid... |

40 |
Generalized additive models: Some applications
- Hastie
- 1987
(Show Context)
Citation Context ... approach that reduces the speci cation of an extremely large contingency table into the speci cation of a few very small tables [13]. These models belong to the more general class of additive models =-=[5, 24, 25]-=- that approximate multivariate functions by sums of univariate functions. Fortuitously, additive belief-network models also overcome the intractability of probabilistic inference, and thus they provid... |

35 |
An introduction to algorithms for inference in belief nets
- Henrion
- 1990
(Show Context)
Citation Context ...ng domain variables [35, 36]. Belief networks are the basis of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks =-=[8, 12, 16, 27, 34]-=- and for speci cation of belief networks [11, 37, 38, 43] have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance... |

27 | A logic and time nets for probabilistic inference
- Kanazawa
- 1991
(Show Context)
Citation Context ...ter, a state-space forecasting model. Since then, few techniques have been developed to address the forecasting problem directly. Several approaches for probabilistic reasoning about change over time =-=[18, 30]-=- and for temporal reasoning with belief networks and in uence diagrams [4, 44, 39] have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices ... |

26 |
A randomized approximation algorithm for probabilistic inference on Bayesian belief networks
- Chavez, Cooper
- 1990
(Show Context)
Citation Context |

25 | Forecasting sleep apnea with dynamic network models - Dagum, Galper - 1993 |

18 | A Bayesian analysis of simulation algorithms for inference in belief networks
- Dagum, Horvitz
- 1993
(Show Context)
Citation Context |

13 |
ARCO1: An application of belief networks to the oil market
- Abramson
- 1991
(Show Context)
Citation Context ... and for temporal reasoning with belief networks and in uence diagrams [4, 44, 39] have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices =-=[1]-=-. We have developed a probability forecasting model through a synthesis of belief-network models and classical time-series analysis [14, 15]. The dynamic network model (DNM) is based 2son the integrat... |

12 |
Temporal Reasoning with Probabilities
- Berzuini, Bellazi, et al.
- 1989
(Show Context)
Citation Context ...loped to address the forecasting problem directly. Several approaches for probabilistic reasoning about change over time [18, 30] and for temporal reasoning with belief networks and in uence diagrams =-=[4, 44, 39]-=- have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices [1]. We have developed a probability forecasting model through a synthesis of beli... |

12 | A statistical semantics for causation
- Pearl, Verma
- 1991
(Show Context)
Citation Context ...of diagnostic systems for many real-world applications [2, 3, 7, 26, 28, 29, 42]. Techniques for probabilistic inference in belief networks [8, 12, 16, 27, 34] and for speci cation of belief networks =-=[11, 37, 38, 43]-=- have emerged in response to the diversity of belief-network applications. More generally, applied statisticians have recognized the importance of graphical structures and the fundamental notion of co... |

9 |
Decision theory in expert systems and arti cial intelligence
- Horvitz, Breese, et al.
- 1988
(Show Context)
Citation Context |

7 |
Additive belief network models
- Dagum, Galper
- 1993
(Show Context)
Citation Context ...series analysis [14, 15]. The dynamic network model (DNM) is based 2son the integration of Bayesian time-series analysis with generalizations of belief-network representation and inference techniques =-=[13]-=-. DNMs inherit the expressiveness of belief networks and are thus suited ideally for domains with categorical variables. The graphical structure of dependencies in DNMs represents the direct causal re... |

6 |
Models for diagnosis and detections of combinations of diseases
- Habbema
- 1976
(Show Context)
Citation Context ...recast models; real-world applications of belief-network models also encounter these problems. These limitations prompted the development of models of causal independence and disjunctive interactions =-=[23, 32, 36]-=- that reduce substantially the storage requirements of conditional probability tables. These models, however, do not address the problem of intractable inference. Elsewhere, we develop additive belief... |

6 |
Combining qualitative and quantitative computation in a ventilator therapy planner
- Rutledge, Thomsen, et al.
- 1989
(Show Context)
Citation Context ...ariables. E ective diagnostic reasoning systems use belief networks to assign probabilities to alternative hypotheses about a patient's health|for example, MUNIN [2], ALARM [3], Path nder [26], VPnet =-=[40]-=-, and QMR-DT [42]|or about the source of failure in complex machinery, including jet engines, electric power generators, and copy machines [28, 29]. These applications of belief networks have focused ... |

6 |
Dynamic programming and in uence diagrams
- Tatman, Shachter
- 1990
(Show Context)
Citation Context ...loped to address the forecasting problem directly. Several approaches for probabilistic reasoning about change over time [18, 30] and for temporal reasoning with belief networks and in uence diagrams =-=[4, 44, 39]-=- have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices [1]. We have developed a probability forecasting model through a synthesis of beli... |

5 |
Bayesian belief-network inference using nested dissection
- Cooper
- 1990
(Show Context)
Citation Context ...ve decomposition to improve the e ciency of inference. The inference algorithm for additive belief networks [13] is similar to Cooper's nested dissection algorithm for general probabilistic inference =-=[9]-=-. The algorithm decomposes the belief network into subnetworks using the additive decompositions. The Lauritzen-Speiglehalter (L-S) algorithm is used [34] to perform inference on the subnetworks. The ... |

3 |
A probabilistic reformulation of the quick medical reference system
- Shwe, Middleton, et al.
- 1990
(Show Context)
Citation Context |

2 |
et al.: APACHE-acute physiology and chronic health evaluations: A physiologically based classification system. Crit Care Med
- Knaus, Zimmerman, et al.
- 1981
(Show Context)
Citation Context ...ne system, the acute physiology and chronic health evaluation (APACHE) system, relates the severity ofa patient's illness to the degree of physiologic derangement of a set of physiologic measurements =-=[33]-=-. Its successor, the APACHE III [20] system, is currently in use at many ICUs to aid in cost-bene t treatment decisions based on the mortality risk predictions of critically ill patients. Other scorin... |

2 |
Tradeo s in constructing and evaluating temporal in uence diagrams
- Provan
(Show Context)
Citation Context ...loped to address the forecasting problem directly. Several approaches for probabilistic reasoning about change over time [18, 30] and for temporal reasoning with belief networks and in uence diagrams =-=[4, 44, 39]-=- have been proposed. Real-world applications of forecasting with belief networks include forecasting crude-oil prices [1]. We have developed a probability forecasting model through a synthesis of beli... |

1 |
Fourth-order blood pressure waves
- Carroll
- 1990
(Show Context)
Citation Context ...s and terminated when the patient was given a rapid intravenous infusion of 500 cc of 6% albumin. discussion The oscillation of pressures, VO2, and SvO2 is an unusual occurrence, rst noted by Carroll =-=[6]-=-. Several other instances of this phenomenon have been discovered subsequently. The pathophysiology of the process has not yet been elucidated. Recent work, however, suggests a link between a sustaine... |

1 |
et al. The APACHE III Prognostic System: Risk prediction of hospital mortality for critically ill hospitalized adults
- William
- 1991
(Show Context)
Citation Context ...chronic health evaluation (APACHE) system, relates the severity ofa patient's illness to the degree of physiologic derangement of a set of physiologic measurements [33]. Its successor, the APACHE III =-=[20]-=- system, is currently in use at many ICUs to aid in cost-bene t treatment decisions based on the mortality risk predictions of critically ill patients. Other scoring schemes have been developed based ... |

1 |
Withdrawing and withholding treatment inintensive care. Part 2. Patient assessment
- Fisher, Raper
- 1990
(Show Context)
Citation Context ...f patient care. Systems to predict patient care would limit admission to the ICU to patients who would clearly bene t from critical care, and would withhold therapy from patients who would not bene t =-=[21]-=-. One system, the acute physiology and chronic health evaluation (APACHE) system, relates the severity ofa patient's illness to the degree of physiologic derangement of a set of physiologic measuremen... |

1 |
uence diagram models with continuous variables
- In
- 1986
(Show Context)
Citation Context ...categorical data [45]. Forecasting has been addressed only intermittently over the last decade by the uncertain reasoning community [19]. Early work on continuous-variable in uence diagrams by Kenley =-=[31]-=- yielded an in uence-diagram representation of the Kalman lter, a state-space forecasting model. Since then, few techniques have been developed to address the forecasting problem directly. Several app... |