Rui Pedro Barbosa

Experimental Platform for Collaborative Inter and Intra Cellular Fuzzy Scheduling in an Ubiquitous Manufacturing System

The aim of manufacturing scheduling is the efficient allocation of machines and other resources to jobs, or operations within jobs, and the subsequent time phasing of these jobs on individual machines. Therefore, the scheduling of production processes of a distributed cellular manufacturing enterprise is one of the significant tasks to be performed to achieve competitive production, which means, e.g., to deliver products on time or to use resources efficiently and reduce production times. In this paper we propose a Web Platform for solving those kind of problems occurring either in intra or inter cellular manufacturing scenarios. Scheduling methods are local or remotely available through web services and can be easily and continuously incorporated in a distributed repository, which integrates XML-based components, belonging to a range of business partners, integrating a Virtual Enterprise, in the context of an Ubiquitous Manufacturing System. The scheduling data modeling and the data transferring processes are based on XML and related web technologies and decision-making is carried out through an interactive approach relying on fuzzy sets and user friendly interfaces for supporting cellular manufacturing scheduling.

Multi-Agent Forex Trading System

Automated trading is a novel field of study in which computer programs are put in charge of deciding when and how to trade financial instruments. Intelligent agents, with their ability to act autonomously and to adapt and interact with the environment, seem like an obvious choice for the development of automated trading systems. The aim of this article is to analyze how well intelligent agents suit this task. We implemented a set of autonomous currency trading agents, using an architecture that consists of an ensemble of classification and regression models, a case-based reasoning system and an expert system. A total of six trading agents were implemented, each being responsible for trading one of the following currency pair in the Forex market: EUR/USD, EUR/JPY, EUR/CHF, USD/JPY, USD/CHF and CHF/JPY. The agents simulated trades over a period of 23 months, having all achieved a reasonable profit trading independently. However, their strategies resulted in relatively high drawdows. In order to decrease the risk inherent to these high drawdowns, the same simulation was performed while making the agents share the monetary resources. As expected, this strategy of investment diversification originated better results. Still, when the trading costs were taken into consideration, the overall trading performance was less than impressive. That was due to the fact that each agent performed too many trades, and the cost associated with the trading commissions became prohibitively high. We were able to lessen the impact of the trading costs in the total profit by integrating the agents in a multi-agent system, in which the agents communicated with each other before opening new trades. This allowed them to calculate the intended exposure to the market, which in turn enabled them to avoid redundant trades. Under simulation and using low leverage, this multi-agent system obtained a 55.7% profit in 23 months of trading, with a 9.0% maximum drawdown.

An Agent Task Force for Stock Trading

In this article the authors present the simulated trading results of a system consisting of 60 intelligent agents, each being responsible for day trading a stock listed on the NYSE or the NASDAQ stock exchange. These agents were implemented according to an architecture that was previously applied to currency trading with interesting results. The performance of the stock trading agents, once integrated in a diversified investment system, showed similar promise. The trading simulation was done using out-of-sample price data for the period between February of 2006 and October of 2010. Throughout this period, the system’s performance compared favorably with that of the buy-and-hold strategy, both in terms of return and maximum drawdown. These results indicate that agent technology might be of use for this particular practical application, a conclusion that should interest the investment industry.

A Step-By-Step Implementation of a Hybrid USD/JPY Trading Agent

In this article we describe the step-by-step implementation of an agent that can trade the USD/JPY currency pair using a 6 hours timeframe. The agent is capable of trading autonomously due to its ability to handle money management and to decide when to buy or sell the currency pair. Its implementation consists of a prediction mechanism, which it uses to forecast the direction of the price, and a risk management system, which enables it to make decisions regarding how much to invest in each trade and when to avoid trading. We present several alternatives for the price prediction mechanism, from using a standalone classification or regression model to using an ensemble with fixed or dynamic vote weights. The agent performed simulated trades over a period of 17 months, and obtained a return of around 50% using low leverage and after taking into account the trading costs.

P2P Web Service Based System for Supporting Decision-Making in Cellular Manufacturing Scheduling

With the increase of the Internet and Virtual Enterprises (VEs), interfaces for web systems and automated services are becoming an emergent necessity. In this paper we propose a Peer-to-peer (P2P) web-based decision-support system for enabling access to different manufacturing scheduling methods, which can be remotely available and accessible from a distributed knowledge base. The XML-based modeling and communication is applied to manufacturing scheduling. Therefore, manufacturing scheduling problems and methods are modeled using XML. The proposed P2P web-based system works as web services, under the SOAP protocol. The system’s distributed knowledge base enables sharing information about scheduling problems and corresponding solving methods in a widened search space, through a scheduling community, integrating a VE. Running several methods enables different results for a given problem, consequently, contributing for a better decision-making. An important aspect is that this knowledge base can be easily and continuously updated by any contributor through the VE. Moreover, through this system once suitable available methods, for a given problem, are identified, it enables running one or more of them, for enabling a better manufacturing scheduling support, enhanced though incorporated fuzzy decision-making procedures.

The Agent-Based Hedge Fund

In this article we describe the implementation of a diversified investment strategy using 25 intelligent agents. Each agent utilizes several data mining models and other artificial intelligence techniques to autonomously day trade an American stock. The agents were individually tested with out-of-sample data corresponding to the period between February of 2006 and June of 2010, and most achieved an acceptable performance. By integrating the 25 agents in a multi-agent system, we were able to obtain much better results (according to the return and maximum drawdown metrics); this leads us to believe that it might be possible to use one such system in the creation of a profitable hedge fund in which the investment decisions can be made without human intervention.

A Diversified Investment Strategy Using Autonomous Agents

In a previously published article, we presented an architecture for implementing agents with the ability to trade autonomously in the Forex market. At the core of this architecture is an ensemble of classification and regression models that is used to predict the direction of the price of a currency pair. In this paper, we will describe a diversified investment strategy consisting of five agents which were implemented using that architecture. By simulating trades with 18 months of out-of-sample data, we will demonstrate that data mining models can produce profitable predictions, and that the trading risk can be diminished through investment diversification.

Lazy classification using an optimized instance-based learner

Classification is a machine learning technique whose objective is the prediction of the class membership of data instances. There are numerous models currently available for performing classification, among which decision trees and artificial neural networks. In this article we describe the implementation of a new lazy classification model called similarity classifier. Given an out-of-sample instance, this model predicts its class by finding the training instances that are similar to it, and returning the most frequent class among these instances. The classifier was implemented using Wekas data mining API, and is available for download. Its performance, according to accuracy and speed metrics, compares relatively well with that of well-established classifiers such as nearest neighbor models or support vector machines. For this reason, the similarity classifier can become a useful instrument in a data mining practitioners tool set.