Forecasting Methods And Stock Market Analysis (Page 5 of 8) in pdf

Forecasting methods and stock market analysis

107

amount of data will be used. Various versions of moving averages are used:

arithmetic moving averages, geometric moving averages, harmonic moving

averages, quadratic moving averages, exponential moving averages, weighted

moving averages, and double moving averages.

The moving average models are certainly useful, but they oﬀer no infor-

mation about the trust in forecast, and do not explain the random behavior

of the time series. The stochastic models are therefore necessary, because

the random component brings information about forecast errors.

The basis for the stochastic modeling is the stochastic process. We

may suppose that a stochastic process generates the time series. The pro-

cess consists of an ordered set of random variables that are associated with

probability distributions, deﬁned for each of the t time moments. We con-

sider that the series of daily shares’ quotations are generated by discrete

stochastic process of real values.

The simplest stochastic process, which generates purely random series,

is known as white noise and consists of a succession of random indepen-

dent variables, identically distributed, and of zero average. A series is a

white noise if it has no known structure (model), and this is why it is non-

predictable.

The random walk process is frequently used as a model for the stock

market quotations. For this particular model all the predictions are equal to

the last observed value, and the conﬁdence intervals are higher as the fore-

cast horizon is expanding. A particular version of this process, the random

walk with tendency, takes into consideration the existence of a tendency and

enables one to include that tendency within the prediction.

The stochastic model of the moving averages thoroughly describes

the y

process as the weighted sum of the actual and future random pertur-

bations.

The knowledge of the properties of the series generated by the moving

average method is achieved by means of the process of series autocorre-

lation. This means the correlation of a series with its own history (y

and

, for example).

t−k

In the autoregressive models y

depends on a weighted sum of the past

values and the term of random perturbation.

There are many random processes for which it is impossible to build a

universally valid model through either the moving average or the purely

autoregressive model. By integrating the autoregressive and moving average

models, the ARMA (M ixed Autoregresive – M oving Average Models)

model was obtained. This is deemed to be the most suitable one for economic

forecasts, when the evolution of the exogenous variables is unknown. In these

models the process y

is a function of both its own past values, on the one

hand, and the past and actual random perturbations, on the other hand.

Forecasting Methods And Stock Market Analysis Page 5

Related Articles

Related forms

Related Categories