Welcome to FutureFusion

FutureFusion combines the concepts of probabilistic reasoning and collaborative knowledge elicitation to compute the probabilities (likelihoods) of predictions based on user feedback. In order to generate the probability of predictions we combine your belief with all other users' belief on:

• a priori likelihoods: The probability of the prediction without the knowledge of the current situation, or in other words the probability the prediction will occur as a result of some unknown influence
• influences on predictions: External influences on a prediction and the strength of the influence (does it decrease or increase the likelihood of a prediction?)

In summary, the beliefs provided by users on a priori likelihood are averaged to find the a priori value. The beliefs in influence strengths are used to determine a conditional likelihood of the prediction given its influence. The conditional likelihood represents the likelihood that an effect (right side of influence) will happen, given that its cause (left side of influence) happens.

Once the average conditional likelihood for each influence is found, the probability of a prediction is determined by aggregating these values with the computed probability of the causes and the probability that the prediction will occur by some unknown cause (the a priori value * the probability that none of the incoming influences will occur).

FutureFusion is based on a rich body of research in computational modeling and artificial intelligence:

• Bayesian Networks: http://en.wikipedia.org/wiki/Bayesian_network
• Bayesian Independence Assumption: http://en.wikipedia.org/wiki/Naive_Bayes_classifier
• Causal Strength Logic: http://www.inet.saic.com/docs/_docs_/math.pdf
• Belief Aggregation: http://www.google.com/search?q=belief+aggregation
• Loopy Belief Propagation: http://en.wikipedia.org/wiki/Belief_propagation