Dynamically unrecognisable factors affecting variations of stock index and lack of intact models for description of each factor and their interactions can be considered as stimuli for forecasting variations of stock index through the methods based on time series of companies. The present paper aimed at forecasting the Tehran securities price index using the nearest neighbour algorithm. The study sample was comprised of some companies accepted in the Tehran stock exchange from 25 March 2006 to 20 March 2010 with 1144 observations. The obtained data were analysed via regression correlation test using statistical software's R and Eviews software's. According to the results and mean squares errors, the nearest neighbour algorithm is more efficient than the ARIMA linear model for forecasting the Tehran securities price index with lower errors.
Forecasting, Price index, Tehran stock exchange, The nearest neighbour algorithm, ARIMA model
