AI has long been applied to the problem of predicting financial markets.

Prediction of financial time series using artificial neural networks has been the subject of many publications, even if the predictability of financial series remains a subject of scientific debate in the financial literature.

this paper is thus to complement the above analyses by considering these questions in the context of analysing a long time series of rather more parochial temperature data. The Central England temperature series was originally constructed by Manley (1974), has been further discussed in Parker, Legg and Folland (1992), and is now routinely updated by the Hadley Centre. The data is available from www.cru.uea.ac.uk/~mikeh/datasets/uk/cet.htm, and comprises a series on monthly mean surface air temperatures for a region representative of the English Midlands, expressed in degrees Celsius for the period from 1659 to present (which for our purposes is the end of 1999). As such it is an excellent vehicle for analysing issues of global warming using time series techniques suitable for extracting trend components.

This paper focuses on the role of technical analysis in signalling the timing of stock market entry and exit. We introduce test statistics to test the performance of the most established of the trend followers, the Moving Average, and the most frequently used counter-trend indicator, the Relative Strength Index. Using Singapore data, the results indicate that the indicators can be used to generate significantly positive return. We find that member firms of Singapore Stock Exchange (SES) tend to enjoy substantial profits by applying technical indicators. This could be the reason why most member firms do have their own trading teams that rely heavily on technical analysis. 1 1

This study tests the performance of stock market forecasts derived from technical analysis by means of a specific indicator. The indicator is computed from E/P ratios and bond yields. Several stock markets are studied over a 20year period. Two test statistics are introduced to utilize the indicator. The results show that the forecasts generated from the indicator would enable investors to escape most of the crashes and catch most of the bull runs. The trading signals provided by the indicator can generate profits that are significantly better than the buy-and-hold strategy (JEL G14, G10). Keywords:bond yield, E/P ratio, interest rate, standardized yield differential, yield differential I.

This paper presents a joint review on professional technical analysis and academic quantitative analysis of the financial markets, aiming at bridging the deep gulf between the two fields and unifying them under a general science of intelligent finance or financial intelligence. While econometricians and econophysicians have recently reexamined technical analysis, most of their effort is focused on chart patterns and technical indicators, leading to some simplicity impression of technical analysis. In our view, the most valuable core and also the hardest part of technical analysis is the fractal and quantum nature of Elliott waves and Gann price-time cycles and angles. On the quantitative analysis side, since Mandelbrot’s discovery of fractals in financial time series, both empirical and fundamental progresses have been made, mainly in the last decade, including a third-order power law asymptotic behavior in return distribution, an accelerated crossover from the power law towards a Gaussian, a theoretical framework of crashes as critical points, and multi-agent game models of the financial markets. Inspired by these developments from the two fields we point out the possibility of developing an adaptive computational model of Elliott waves and Gann price-time cycles and angles using multilevel power laws, log-periodicity and instantaneous phase estimation.

We present an artificial stock market in which simple trading agents enter an asynchronous double auction market to trade in a single stock. Beginning with a population of random trading agents drawing their bid prices from a normal distribution around the current price, we compare the statistical properties of the emergent stock price return distribution to that observed on a real price series, the daily returns distribution for GE stock listed on the NYSE.

Embedded computing devices that interact with humans and the real world hold great promise in making our lives more comfortable and convenient – perhaps allowing independence longer and later in life, or better understand the changes in our natural environment. The biggest difficulty in taking advantage of these computers is that they need too much assistance from us, starting with configuration, with adapting to new dynamic requirements, and ending in learning from our intent. The ubiquity of computers makes the situation only worse— telling all the little computers what to do is easily harder than simply doing the task yourself. We claim that the only way to make ubiquitous computing a practicality is to enable the computers to figure out what to do on their own. Observe and learn, or perish. This paper proposes a framework for automatic configuration and adaptation using learning and prediction based on observed context histories. A software architecture for describing, recording, analyzing, and reacting to physical or computational variables is substantiated with a case study that self-tunes distributed real-time tasks in an entertainment scenario. The measured results are generalized, using stochastic or physical models, to apply to a large number of problems that allow ubiquitous computing to become a reality.

Abstract. This paper provides an analytical approach to fuzzy rule base optimization. While most research in the area has been done experimentally, our theoretical considerations give new insights to the task. Using the symmetry that is inherent in our formulation, we show that the problem of finding an optimal rule base can be reduced to solving a set of quadratic equations that generically have a one dimensional solution space. This alternate problem specification can enable new approaches for rule base optimization. 1

This paper analyzes the relation between momentum strategies (strategies that buy stocks with high returns over the previous three to 12 months and sell stocks with low returns over the same period) and turnover (number of shares traded divided by the number of shares outstanding) for the German stock market. Our main finding is that momentum strategies are more profitable among highturnover stocks. We present various robustness checks, long-horizon results, evidence on seasonality, and control for size-, book-to-market-, and industry-effects. We argue that our results are useful to empirically evaluate competing explanations for the momentum effect.