Nigel Meade(2002) A comparison of the accuracy of short term foreign exchange forecasting methods. International Journal of Forecasting, Vol. 18, Iss. 1; pg. 67 Abstract: Evidence for a non-linear generating process is evaluated by an analysis of the comparative accuracy of short term forecasts of foreign exchange (FX) rates. Forecasts were generated by a linear AR-GARCH model and four non-linear methods, including three nearest neighbour methods and locally weighted regression. Five data frequencies were used: daily, four-hourly, two-hourly, hourly and half-hourly. Using root mean square error as a measure, significantly greater accuracy than a no-change forecast was achieved for two-hourly and higher frequency data sets. Using a test by Peseran and Timmerman, significant predictive directional accuracy was found for four-hourly and higher frequency data sets. These results were supported by simulated trading based on forecast direction. No evidence was found that the FX rate behaviour is better represented by a non-linear generating process than by a linear model. Ramazan Gencay(1999) Linear, non-linear and essential foreign exchange rate prediction with simple technical trading rules. Journal of International Economics, Vol. 47, Iss. 1; pg. 91 Abstract: A paper investigates the predictability of spot foreign exchange rate returns from past buy-sell signals of the simple technical trading rules by using the nearest neighbors and the feedforward network regressions. The optimal choices for nearest neighbors, hidden units in a feedforward network and the training set are determined by the cross validation method which minimizes the mean square error. Although this method is computationally expensive, the results indicate that it has the advantage of avoiding overfitting in noisy environments and indicate that simple technical rules provide significant forecast improvements for the current returns over the random walk model. Mizrach, B(1992) Multivariate nearest-neighbour forecasts of EMS exchange rates. Journal Of Applied Econometrics, Vol. 7; pg. S151, 13 pgs Abstract: Exchange rate modeling has been a persistent puzzle for international economists. Forecasts from popular models for the exchange rate generally fail to improve upon the random walk out-of-sample. While a multivariate nonparametric approach provides useful information about exchange rates, the model produces forecasts superior to the random walk for only one of the 3 European monetary system currencies examined. Using a statistic developed in Mizrach (1991), the forecast improvement found - a 4%-5% reduction in mean squared error for the lira in daily returns - is not statistically significant. A cross-validation exercise suggests that the improvement is also not robust. LeBaron, B(1992) Forecast improvements using a volatility index. Journal Of Applied Econometrics, Vol. 7; pg. S137, 13 pgs Abstract: The possibility of improved out of sample forecasting for stock returns and foreign exchange rates are explored using observed nonlinearities in the 2 series. Forecasting is done using nonparametric techniques where important information is obtained from the current level of volatility in the series. Forecast improvements are observed for both series, but for stock returns, the improvements are only marginal. These results indicate the usefulness and stability of some types of nonlinear modeling for financial markets. Through this analysis, 2 important points about forecasting can be concluded: 1. When improvements are measured is more important than how forecasting is done. 2. There may be simple modifications to the standard nearest-neighbor techniques, which will yield significant improvements. ----------------------------- Abhyankar, A(1997) Uncovering nonlinear structure in real-time stock-market indexes: The S&P 500, the DAX, the Nikkei 225, and the FTSE-100. Journal of Business & Economic Statistics, Vol. 15, Iss. 1; pg. 1, 14 pgs Abstract: Nonlinear dependence and chaos in real-time returns are tested on the world's 4 most important stock-market indexes. Both the Brock-Dechert-Scheinkman (1987) and the Lee, White and Granger (1993) neural-network-based tests indicate persistent nonlinear structure in the series. Estimates of the Lyapunov exponents using the Nychka, Ellner, Gallant and McCaffrey (1992) neural-net method and the Zeng, Pielke and Eyckholt (1992) nearest-neighbor algorithm confirm the presence of nonlinear dependence in the returns on all indexes but provide no evidence of low-dimensional chaotic processes. Given the sensitivity of the results to the estimation parameters, it is concluded that the data are dominated by a stochastic component. ----------------------------- Andrey D Pavlov(2000) Space-varying regression coefficients: A semi-parametric approach applied to real estate markets. Real Estate Economics, Vol. 28, Iss. 2; pg. 249, 35 pgs Abstract: This paper presents a method for estimating home values by non-parametrically incorporating the physical location of the properties. Specifically, the parameters of the observed covariates are allowed to vary in space. This approach mitigates one of the biggest deficiencies inherent in hedonic pricing models - omitted variables. The advantages of the proposed method are demonstrated using real estate transaction data from Los Angeles County. The estimation finds a substantial spatial variation of the marginal values of the hedonic characteristics and provides an insight into the segmentation of the market. The proposed method is an extension of semi-parametric multi-dimensional k-nearest-neighbor smoothing. It alleviates a fundamental problem known as the curse of dimensionality by incorporating parametric components into a non-parametric estimation.