Collaborating Authors


AAAI Conferences

Given a set of training examples S and a treestructured attribute X, the goal in this work is to find a multiple-split test defined on x that maximizes Quinlan's gain-ratio measure. The number of possible such multiple-split tests grows exponentially in the size of the hierarchy associated with the attribute. It is, therefore, impractical to enumerate and evaluate all these tests in order to choose the best one. We introduce an efficient algorithm for solving this problem that guarantees maximizing the gain-ratio over all possible tests.

Early prediction of electric power system blackouts by temporal machine learning

AAAI Conferences

This paper discusses the application of machine learning to the design of power system blackout prediction criteria, using a large data base of random power system scenarios generated by Monte-Carlo simulation. Each scenario is described by temporal variables and sequences of events describing the dynamics of the system as it might be observed from realtime measurements. The aim is to exploit the data base in order to derive as simple as possible rules which would allow to detect an incipient blackout early enough to prevent or mitigate it. We propose a novel "temporal tree induction" algorithm in order to exploit temporal attributes and reach a compromise between degree of anticipation and selectivity of detection rules. Tests are carried out on a data base related to voltage collapse of an existing large scale power system.

Avoiding Bias when Aggregating Relational Data with Degree Disparity

AAAI Conferences

A common characteristic of relational data sets --degree disparity--can lead relational learning algorithms to discover misleading correlations. Degree disparity occurs when the frequency of a relation is correlated with the values of the target variable. In such cases, aggregation functions used by many relational learning algorithms will result in misleading correlations and added complexity in models. We examine this problem through a combination of simulations and experiments. We show how two novel hypothesis testing procedures can adjust for the effects of using aggregation functions in the presence of degree disparity.

The Complete Guide to Decision Trees (part 2)


Now you may ask yourself: how do DTs know which features to select and how to split the data? To understand that, we need to get into some details. All DTs perform basically the same task: they examine all the attributes of the dataset to find the ones that give the best possible result by splitting the data into subgroups. They perform this task recursively by splitting subgroups into smaller and smaller units until the Tree is finished (stopped by certain criteria). This decision of making splits heavily affects the Tree's accuracy and performance, and for that decision, DTs can use different algorithms that differ in the possible structure of the Tree (e.g. the number of splits per node), the criteria on how to perform the splits, and when to stop splitting.

Comparison of data mining techniques and tools for data classification (PDF Download Available)


The datasets used in the test were saved in Weka's standardized All tools are able to read this format natively. It was not used any preprocessing widget. Bayes', the data has not been subjected to preprocessing as these'RProp MLP Learner', the real class and the predicted class were Tests were exhaustive, i.e. all the algorithms were RapidMiner has some operators (e.g. 'LibSVMLearner'), that only work with numeric attributes; for