Hi Manas, when you try to forecast a time series, you can use fitness functions based on distance (MSE, SSE, ASE, ...), variability (MAD, STD, ...), and information criterias (AKAIKE - AIC, BIC, SRM ...), depend of you necessities.
Hi Manas. the moving average is actually not a "fitness function" (i.e. loss function, cost of error function, objective function) but actually an algorithm to predict the future realisations of said time series by smoothing out randomness.
Are you looking for algorithms to predict S&P values into the future? Or are you looking for a fitness function to determine the value (or rather loss or errors) of your predictions?
Hi Manas. Sorry, but I dont quit eunderstand. I think you need to elaborate a bit more and be much more specific as the sentence you wrote does not really make sense to me. What is the input data you want to use: S&P prices, returns, hi low open close, ... ?Do you want to predict the price level of the S&P (i.e. an interval scaled variable)? Do you want to predict the direction of the price change to S&P from one period to the next (i.e. a classification problem)? Do you want to predict a quantile of the returns? Do you want to predict the volatility of the prices or returns? ...This would make it easier to help you. Thx for clarifying. Sven
I have a data in which attributes are Closing price, opening price and high price and no. of volumes traded based on these attributes i want to predict whether the stock will show upwards movement or downward movement. For that i need fitness function to select those tuples that possess minimum outliers and provide effective prediction