Pier Paolo Puliafito

An Application of Data Mining to the Problem of the University Students' Dropout Using Markov Chains

A new application of data mining to the problem of University dropout is presented. A new modeling technique, based on Markov chains, has been developed to mine information from data about the University students’ behavior. The information extracted by means of the proposed technique has been used to deeply understand the dropout problem, to find out the high-risk population and to drive the design of suitable politics to reduce it. To represent the behavior of the students the available data have been modeled as a Markov chain and the associated transition probabilities have been used as a base to extract the aforesaid behavioral patterns. The developed technique is general and can be successfully used to study a large range of decisional problems dealing with data in the form of events or time series. The results of the application of the proposed technique to the students’ data will be presented.

Some problems on uncertain knowledge acquisition for rule based systems

Abstract The problem of uncertainty management in systems based on rules can be dealt with using several methodologies. A particularly significant and well-founded one is based on Dempster-Shafers Theory of Evidence. This methodology is based on the hypothesis that the values of an attribute for a certain object are mutually exclusive. This hypothesis is often too restrictive and the present work proposes a procedure of knowledge acquisition which allows for the application of the D-S theory even in cases where the objects are described, for example, in a liguistically ‘conventional’ manner, where such a constraint would not be satisfied. This paper introduces the concept of knoxel and relates also to other problems which are of particular interest for the use of complex networks based on knoxels.

A Combinatorial Method for Health-Care Districting

The problem described in this paper arises whenever, for political or administrative reasons, a geographical region must be partitioned into an unknown number of districts so that, within each district,the total amount of service supplied by existing facilities of known location and capacity may satisfy a given demand, according to some optimality criterion. The philosophy of the problem is quite different from that of the so-called Location Problem (see for example TOREGAS et AL. [I]), since we are neither seeking locations for new facilities nor trying to modify their capacities. The following assumptions will define the problem more clearly. I. The region is considered to be made up of a certain number of assigned undivisible communities of known concentrated population. 2. Each community must be univocally assigned to a district, no overlap among districts is allowed. 3. The location and capacity of every facility are assigned. 4. 5. 6. i. e. ,

A New Modeling Technique Based on Markov Chains to Mine Behavioral Patterns in Event Based Time Series

A new modeling technique to mine information from data that are expressed in the form of events associated to entities is presented. In particular such a technique aims at extracting non-evident behavioral patterns from data in order to identify different classes of entities in the considered population. To represent the behavior of the entities a Markov chain model is adopted and the transition probabilities for such a model are computed. The information extracted by means of the proposed technique can be used as decisional support in a large range of problems, such as marketing or social behavioral questions. A case study concerning the university dropout problem is presented together with further development of Markov chain modeling technique in order to improve the prediction and/or interpretation power.

Optimal Control of Data Flow from Peripheral to Central Units in an Information Structure

Abstract The optimal control of data flow from peripheral units to a central one is considered. A central unit exerts control actions on a linear dynamic system. Communications are costly and the performance criterion of the process control is quadratic. Task centralization in selecting communication instants is compared with task decentralization. The latter case is approached by means of team theory, and necessary conditions for optimality are established.

Optimization of costly measurements in stochastic decision processes

SummaryTwo kinds of problems regarding measurement optimization in stochastic decision processes, when measurements are costly or constrained not to exceed a given number, have been investigated in the last years: the first one refers to the optimum timing of observations on the state vector of the process, while the second refers to the convenience of buying information on the random actions exerted by a stochastic environment. In this paper the two problems are considered from a unified point of view. In other words, the decision maker has to determine the optimal observation policy, under the assumption that both state and random vectors are measurable. A solution based on the application of dynamic programming is discussed for a general class of multistage processes. Analytical results are then obtained for scalar linear systems with quadratic cost on state and control. In this case, an efficient algorithm to determine the optimal sequence of measurements is presented, on the base of which a sensitivity analysis with respect to the process parameters is also carried out.ZusammenfassungIn den letzten Jahren wurden zwei Typen von Problemen untersucht, die sich auf die Optimierung von Messungen bei stochastischen Entscheidungsprozessen beziehen, einmal der Fall, daß Messungen kostspielig sind, und zum anderen, daß deren Anzahl beschränkt ist. Der erste Fall bezieht sich auf die Gewinnung von Informationen über den stochastischen Störprozeß, der zweite Fall untersucht den optimalen Zeitabstand zwischen den Messungen des Zustands- (bzw. Ist-)-vektors. In diesem Beitrag wurden beide Probleme von einem gemeinsamen Gesichtspunkt aus betrachtet, d. h. der Entscheidungsträger soll die optimale Meßpolitik bestimmen, unter der Annahme, daß Zustands- und Störvektor meßbar sind. Mit Hilfe des Dynamischen Programmierens wird eine allgemeine Lösung des Problems hergeleitet. Sodann werden analytische Ergebnisse für den Fall linear-quadratischer Modelle angegeben. Schließlich wird ein wirksamer Algorithmus zur Bestimmung der optimalen Meßfolge entwickelt und darüber hinaus eine Sensitivitätsanalyse der Abhängigkeit von den Prozeßparametern durchgeführt.

Modelling for the Optimal Sizing of an Automatic Intermodal Freight Terminal

Abstract In this paper the problem of the strategic design of an automatic intermodal terminal is addressed. In particular, the considered terminal is supposed to be provided with an innovative transfer system (such as the Italian system Metrocargo) allowing to load/unload containers in a fast and horizontal way under the electric line. In this system, the main strategic decisions regard the number of shuttles, as well as the number of storage locations to be used in the terminal. At this purpose, an optimization problem is stated for determining the optimal resources in order to minimize the overall cost and taking into account some specific constraints. An algorithm is also defined for solving the proposed optimization problem.