Tommaso Minerva

Clustering financial time series: an application to mutual funds style analysis

Abstract Classification can be useful in giving a synthetic and informative description of contexts characterized by high degrees of complexity. Different approaches could be adopted to tackle the classification problem: statistical tools may contribute to increase the degree of confidence in the classification scheme. A classification algorithm for mutual funds style analysis is proposed, which combines different statistical techniques and exploits information readily available at low cost. Objective, representative, consistent and empirically testable classification schemes are strongly sought for in this field in order to give reliable information to investors and fund managers who are interested in evaluating and comparing different financial products. Institutional classification schemes, when available, do not always provide consistent and representative peer groups of funds. A “return-based” classification scheme is proposed, which aims at identifying mutual funds’ styles by analysing time series of past returns. The proposed classification procedure consists of three basic steps: (a) a dimensionality reduction step based on principal component analysis, (b) a clustering step that exploits a robust evolutionary clustering methodology, and (c) a style identification step via a constrained regression model first proposed by William Sharpe. The algorithm is tested on a sample of Italian mutual funds and achieves satisfactory results with respect to (i) the agreement with the existing institutional classification and (ii) the explanatory power of out of sample variability in the cross-section of returns.

Evolutionary Approaches for Cluster Analysis

The determination of the number of groups in a dataset, their composition and the most relevant measurements to be considered in clustering the data, is a high-demanding task, especially when the a priori information on the dataset is limited. Three different genetic approaches are introduced in this paper as tools for automatic data clustering and features selection. They differ in the adopted codification of the grouping problem, not in the evolutionary operator and parameters. Two of them deals with the grouping problem in a deterministic framework. The first directly approaches the grouping problem as a combinatorial one. The second tries to determine some relevant points in the data domain to be used in clustering data as group separators. A probabilistic framework is then introduced with the third one, which starts specifying the statistical model from which data are assumed to be drawn. The evolutionary approaches are, finally, compared with respect to classical partitional clustering algorithms on simulated data and on Fisher’s Iris dataset used as a benchmark.

Building ARMA Models with Genetic Algorithms

The current state of the art in selecting ARMA time series models requires competence and experience on the part of the practitioner, and sometimes the results are not very satisfactory. In this paper, we propose a new automatic approach to the model selection problem, based upon evolutionary computation. We build a genetic algorithm which evolves the representation of a predictive model, choosing both the orders and the predictors of the model. In simulation studies, the procedure succeeded in identifying the data generating process in the great majority of cases studied.

Surface effects on the electronic properties of YBa2Cu3O7

We present the results of a theoretical study of the modifications induced by the surface on the electronic structure of YBa2Cu3O7. Basal plane surfaces terminated with Ba and with Cu-O planes are considered. The calculated local densities of states show that the main structures of the bulk density of states may be drastically modified, specially near the Fermi energy, and new surface derived features appear in the spectrum. Significant changes occur also in the occupancies of the atoms near the surface.

Evolutionary approaches for statistical modelling

In this paper, we describe some evolutionary approaches based on genetic algorithms to deal with the statistical model selection problem using completely data-driven algorithms. First, we propose an approach to select multivariate linear regression models as well as to build ARMA time-series models. Then we introduce a methodology to tackle the clustering problem in a model-based framework. We report the results from several applications and from simulated data sets, and we compare the evolutionary approaches with some classical ones.

Theoretical interpretation of valence band photoemission spectra in YBa2Cu3O7

Abstract Angle resolved one-hole spectra have been calculated starting from a realistic tight-binding Hamiltonian and including correlation and surface effects. The hole self-energy has been evaluated using the t -matrix approximation for the hole-hole scattering. To account for surface effects the local density of states of the outermost atomic layers has been determined assuming an ideal termination of the crystal and a non polar basal plane surface. The theoretical curves allow one to assign the main features of the spectra at high symmetry points of the Brillouin zone, except for the peak observed at 8.5 eV binding energy. It is suggested that the discrepency may arise from the neglecting of the k -dependence in the self-energy or from the extrinsic nature of the peak.

A Neural Net Model to Predict High Tides in Venice

In this research we design and apply a neural network model to predict the tidal levels in the Venetian lagoon. We use an evolutionary computational approach to select the net topology within the class of multilayered feedforward networks. We build a genetic algorithm, which evolves both the number of predictors and the best set of predictors for the model. The results of this approach are compared to the results we achieve with a linear model based on the same set of candidate predictor variables, whose specification is also obtained with a genetic algorithm. The predictions resulting from the genetically evolved neural net model are more accurate for both tidal levels and extreme values (“the high waters”).

Advances in statistical models for data analysis

This paper aims to propose an innovative approach to identify a typology in a quantile regression model. Quantile regression is a regression technique that allows to focus on the effects that a set of explanatory variables has on the entire conditional distribution of a dependent variable. The proposal concerns the use of multivariate techniques to simultaneously cluster and model data and it is illustrated using an empirical analysis. This analysis regards the impact of student features on the university outcome, measured by the degree mark. The analysis is based on the idea that the dependence structure could be different for units belonging to different groups.

OXYGEN-INDUCED SURFACE-STATES IN YBA2CU3O7

Abstract We present the results of theoretical calculations of the surface electronic structure of YBa2Cu3C7 assuming different crystal terminations. Through a detailed analysis of the distribution of the valence charge near the surface we show that hole distribution may be significantly modified at the surface, due to the band narrowing and to the presence of surface states. The role of these modifications in determining the shape of the one-hole spectra is discussed.

One hole spectra at YBa2Cu3O7 surfaces

Abstract We present a theoretical study of the modification of the one-hole spectra induced by the surface in YBa 2 Cu 3 O 7 . Basal plane surfaces terminated either with Ba or CuO planes are considered. Correlation effects are included by using a Hubbard model hamiltonian and by calculating the self-energy in the low density approximation. The results indicate that both the main bands and the satellites are sensitive to the choice of the surface.