Luigi Troiano

A Fast Algorithm for Mining Rare Itemsets

Mining patterns in large databases is a challenging task facing NP-hard problems. Research focused attention on the most occurrent patterns, although less frequent patterns still offer interesting insights. In this paper we propose a new algorithm for discovering infrequent patterns and compare it to other solutions.

Supporting trading strategies by inverse fuzzy transform

Trading in finance requires to define strategies able to identify early trading buy/sell signals, i.e., conditions which suggest to enter or exit a position, in order to exploit early mover advantage. Trading signals are generally identified by looking at the time series of prices and volumes. Several technical analysis indicators and strategies have been proposed and are commonly in use. In this paper we propose inverse fuzzy transform as a means for building a new class of technical indicators. Experimental results show that this approach outperforms simple and exponential moving average when embedded in common strategies.

Adapting palettes to color vision deficiencies by genetic algorithm

In choosing a color palette, it is necessary to take into the account the needs of color vision impaired users, in order to make information and services accessible to a broader audience. This means to search the space of color palettes aimed to find a color combination representing a good trade-off between aesthetics and accessibility requirements. In this paper, we present a solution based on genetic algorithms. Experimental results highlight this approach to be an efficient but effective way to assist UI designers by suggesting appropriate variations of color palettes.

Recursive and iterative OWA operators

An important issue when using the OWA aggregation operators is the determination of weights. One approach is to link the weights to a desired attitudinal character for the aggregation. The ME-OWA operators provide a pioneering example of this approach. Here we first present an alternative approach to generating OWA weights with a desired attitudinal character. We accomplish this by using a family of recursive OWA operators (R-OWA). We then generalize this with a class that allows of OWA aggregation by iteration (It-OWA). Both families are built with the constraint of keeping constant the attitudinal character at any recursion or any iteration step. This is particularly useful in aggregations that sequentially add arguments to the aggregation.

A time-efficient breadth-first level-wise lattice-traversal algorithm to discover rare itemsets

In this paper we face the problem of searching for rare itemsets. A main issue regards the strategy to adopt in exploring the power set lattice. Assuming a power set lattice with full set at the top and empty set at the bottom, the most of the algorithms adopt a bottom-up exploration, i.e. moving from smaller to larger sets. Although this approach is advantageous in the case of frequent itemsets, it might not be worth being used for rare itemsets, as they occur on the top of the lattice. We propose Rarity, a top-down breadth-first level-wise algorithm. Experimental results and comparisons are illustrated in order to provide a quantitative characterization of algorithm performances and complexity. Application to some UCI benchmark and real world datasets is provided. An algorithm parallelization is outlined. Experiments showed that this approach takes advantage of finding all rare non-zero itemsets in less time than other solutions, at expenses of higher memory demand.

Mining frequent itemsets in data streams within a time horizon

In this paper, we present an algorithm for mining frequent itemsets in a stream of transactions within a limited time horizon. In contrast to other approaches that are presented in the literature, the proposed algorithm makes use of a test window that can discard non-frequent itemsets from a set of candidates. The efficiency of this approach relies on the property that the higher the support threshold is, the smaller the test window is. In addition to considering a sharp horizon, we consider a smooth window. Indeed, in many applications that are of practical interest, not all of the time slots have the same relevance, e.g., more recent slots can be more interesting than older slots. Smoothness can be determined in both qualitative and quantitative terms. A comparison to other algorithms is conducted. The experimental results prove that the proposed solution is faster than other approaches but has a slightly higher cost in terms of memory.

Texture recognition by using GLCM and various aggregation functions

We discuss the problem of texture recognition based on the grey level co-occurrence matrix (GLCM). We performed a number of numerical experiments to establish whether the accuracy of classification is optimal when GLCM entries are aggregated into standard metrics like contrast, dissimilarity, homogeneity, entropy, etc., and compared these metrics to several alternative aggregation methods.We conclude that k nearest neighbors classification based on raw GLCM entries typically works better than classification based on the standard metrics for noiseless data, that metrics based on principal component analysis inprove classification, and that a simple change from the arithmetic to quadratic mean in calculating the standard metrics also improves classification.

The importance of dealing with uncertainty in the evaluation of software engineering methods and tools

The correct choice of software tools and methods is a critical success factor to reach and maintain market leadership. A mature approach to estimate the impact and risk of technology adoption is required. This paper underlines the need for dealing with uncertainty to manage correctly the risk of decision-making and proposes a method for evaluating software engineering methods and tools. The method, named Software Engineering Fuzzy Evaluation Method (SEFEM) is centred on a new class of fuzzy aggregators named Ordered Fuzzy Number Weighted Averaging (OFNWA).

Neural Network aided Glitch-Burst Discrimination and Glitch Classification

We investigate the potential of neural-network based classifiers for discriminating gravitational wave bursts (GWBs) of a given canonical family (e.g. core-collapse supernova waveforms) from typical transient instrumental artifacts (glitches), in the data of a single detector. The further classification of glitches into typical sets is explored.In order to provide a proof of concept,we use the core-collapse supernova waveform catalog produced by H. Dimmelmeier and co-Workers, and the data base of glitches observed in laser interferometer gravitational wave observatory (LIGO) data maintained by P. Saulson and co-Workers to construct datasets of (windowed) transient waveforms (glitches and bursts) in additive (Gaussian and compound-Gaussian) noise with different signal-tonoise ratios (SNR). Principal component analysis (PCA) is next implemented for reducing data dimensionality, yielding results consistent with, and extending those in the literature. Then, a multilayer perceptron is trained by a backpropagation algorithm (MLP-BP) on a data subset, and used to classify the transients as glitch or burst. A Self-Organizing Map (SOM) architecture is finally used to classify the glitches. The glitch/burst discrimination and glitch classification abilities are gauged in terms of the related truth tables. Preliminary results suggest that the approach is effective and robust throughout the SNR range of practical interest. Perspective applications pertain both to distributed (network, multisensor) detection of GWBs, where someintelligenceat the single node level can be introduced, and instrument diagnostics/optimization, where spurious transients can be identified, classified and hopefully traced back to their entry points

Interpretability of fuzzy association rules as means of discovering threats to privacy

This paper focuses on studying how data privacy could be preserved with fuzzy rule bases as interpretable as possible. These fuzzy rule bases are obtained from a data mining strategy based on building a decision tree. The antecedents of each rule produced by these systems contain information about the released variables (quasi-identifier), whereas the consequent contains information only about the protected variable. Experimental results show that fuzzy rules are generally simpler and easier to interpret than other approaches but the risk of disclosing does not increase.