What forecasting models for stock n share?

[fishbuff]

What kind do analysts use?arima? Tbat? Stl?ANN?

Sent from smlj using GAGT

 

[Darkzi0n]

if the stock price evolve according to random walk, then it cant be predicted.

most, if not all, time series models work by assuming past pattern repeat itself.... which is definitely not a characteristic of stock prices.

not sure about ANN tho... but from someone without deep knowledge in ANN, my question is how do u find relevant signals to feed into the model, while miraculously filtering out spurious relationship.

perhaps u can try looking out for cointergrated series.

 

[Maeda_Toshiie]

if the stock price evolve according to random walk, then it cant be predicted.

most, if not all, time series models work by assuming past pattern repeat itself.... which is definitely not a characteristic of stock prices.

not sure about ANN tho... but from someone without deep knowledge in ANN, my question is how do u find relevant signals to feed into the model, while miraculously filtering out spurious relationship.

perhaps u can try looking out for cointergrated series.

ANNs are just hugely over parameterized models built from a massive number of simple equations. There is no intelligence in them; they are just (maybe) better at learning patterns for prediction purposes.

 

[wahkao3]

i even thought of using complex stuff like IRR, NPV, DCF, linear regressions, monte carlo simulations


in the end, plain and simple TA+FA is still the best. no nid what complicated model.

 

[fishbuff]

i dont know anything about stocks as i do mostly on time series especially electricity with seasonality and trend, either is additive or multiplicative with either non, semi or full parametric sets from external factors like temperature, humdity, workdays/weekend,s holidays, economic conditions etc.

below is an example of using arima to forecast the future.

those models like arima and exponential trend smoothing are usually backward looking. NNAR can be used to do forecasting too. here's one example;

well, usually we can measure the accuracy by their residuals and absolute errors like MSE, MAPE etc...

i think these forecasting models are not suitable for stocks or shares prediction as they are more volatile than these well defined time series. am i correct to say that?

 

[wahkao3]

i dont know anything about stocks as i do mostly on time series especially electricity with seasonality and trend, either is additive or multiplicative with either non, semi or full parametric sets from external factors like temperature, humdity, workdays/weekend,s holidays, economic conditions etc.

below is an example of using arima to forecast the future.

those models like arima and exponential trend smoothing are usually backward looking. NNAR can be used to do forecasting too. here's one example;

well, usually we can measure the accuracy by their residuals and absolute errors like MSE, MAPE etc...

i think these forecasting models are not suitable for stocks or shares prediction as they are more volatile than these well defined time series. am i correct to say that?

wa so complex!
this one must ask our inhouse quant expert, shiny things liao.
paging for shiny things

 

[alexchia01]

i dont know anything about stocks as i do mostly on time series especially electricity with seasonality and trend, either is additive or multiplicative with either non, semi or full parametric sets from external factors like temperature, humdity, workdays/weekend,s holidays, economic conditions etc.

below is an example of using arima to forecast the future.

those models like arima and exponential trend smoothing are usually backward looking. NNAR can be used to do forecasting too. here's one example;

well, usually we can measure the accuracy by their residuals and absolute errors like MSE, MAPE etc...

i think these forecasting models are not suitable for stocks or shares prediction as they are more volatile than these well defined time series. am i correct to say that?

Deed but impractical

 

[wahkao3]

Dr Tee optimize and perrisism techniquie leh? i think maybe can use