System for analysis and prediction of financial and statistical data

Abstract

System for the analysis and prediction of the evolution of various types of data such as stock market prices, financial indices and other statistical data, based on the presence of characteristic figures in a dense network of curves constructed from data.

Description

FIELD OF INVENTION

The present invention relates to the analysis and prediction of the evolution of various types of data such as stock market prices, financial indices and other statistical data.

BACKGROUND OF THE INVENTION

Currently, the following two methods are used to analyze and predict data in the field of finance and economics:

• • Technical analysis, based exclusively on the examination of a small number of technical indicators derived from the given data;
  • Fundamental analysis, based on knowledge of the economic situation with regard to the data considered.
    These two approaches often result in predictions that not only differ, but are also often invalidated afterward.

SUMMARY OF THE INVENTION

The present system allows for a superior level of analysis and prediction of the evolution of the aforementioned data, both qualitatively and quantitatively. It rests primarily on a dense network of curves constructed mathematically from numerical data (for example, a stock price) and defined by a primary parameter (the number of data points used) and a secondary parameter (the scale parameter). A computer is used to receive and process the data.

The curves of this network belong to one of the following categories:

• Moving regression (MR) of degree zero, known as the moving average (MA);
• MR of the first degree, known as the moving linear regression (MLR);
• MR of the second degree, which we will call the moving quadratic regression (MQR);
• MR of the k^th degree, which we will call the moving k regression (MKR).

The MA is a well-known indicator commonly used in technical analysis. The MLR, a known but seldom used technical indicator, is built upon the linear regression according to a defined method. The MQR is built upon the quadratic regression according to the same method. The MKR is built similarly upon a regression of the k^th degree.

The present system is fundamentally based on the utilization of a dense network of MRs corresponding to a large set of values of the primary parameter, chosen according to defined criteria.

When MLRs are used to construct the dense network, characteristic figures appear strikingly on the monitor of a computer. For this reason and others that will be discussed later, the network described in what follows is composed of MLRs. It is on the presence of these characteristic figures within the dense network that rests the ability to obtain precise and reliable information on the evolution of the data under consideration.

The system can also use adjusted data, for example, averaged or weighted data.

The secondary parameter (the scale parameter) can be the interval of time separating two consecutive data points, for example, minutes, hours or days. Other types of intervals can also be used; for a financial market, for example, the interval can be expressed in terms of the number of exchanges.

The necessary conditions under which the characteristic figures appear in the network are the following:

• 1) The network must contain a large number of MLRs, greater than about 20. For these characteristic figures to be better observed, ideally, this number must be greater than 100;
• 2) The set of the values of the primary parameter must extend over a sufficiently large range;
• 3) The distribution of the values of the primary parameter must be such that the corresponding network has a uniform density on average.

In practice, criterion 3) is satisfied when the values of the primary parameter constituting the set grow slowly and uniformly. Furthermore, if wished, one can slightly modify the density, for example, by making the network denser for smaller values of the primary parameter.

The following algebraic formula is used to determine with more than sufficient precision the values of the primary parameter, including the possibility of modifying the density: $n_k=n_1+\left(k-1\right)a+\frac{k\left(k-1\right)}{N\left(N-1\right)}\left[n_N-n_1-\left(N-1\right)a\right]$
where:

• k ={1, . . . N};
• N is the number of curves in the network;
• n₁ is the first term of the set;
• n_N is the N^th term of the set; and
• a is the interval between n₁ and n₂.

Taking N =100, n₁=8, n_N=1502, and a =8 as an example, one obtains for the primary parameter the following set of values:

• {8, 16, 24, 33, 41, 50, 59, 68, . . . , 1351, 1372, 1393, 1415, 1436, 1458, 1480, 1502} This set of values generates a network of 100 MLRs which, as desired, has a uniform density on average and extends over a large range.

The characteristic figures seen on the monitor of the computer belong to one of the following three types:

• 1) Cords;
• 2) Envelopes;
• 3) Boltropes.

A cord is a pronounced condensation of curves that stands out from a less dense background of curves of the network.

An envelope outlines the boundary of a group of curves of the network.

A boltrope is both a cord and an envelope.

A characteristic figure attracts or repels the representative curve of the data, depending on its type, its shape and its relative position to the representative curve of the data. The more marked the characteristic figure, the stronger the attraction or the repulsion.

The analysis and prediction of the evolution of the data requires the examination of the ensemble of the cords, envelopes and boltropes and the representative curve of the data up to a given moment, over a sufficiently large interval of consecutive data points. An interval is considered sufficiently large when it contains a peripheral characteristic figure at the top of the network exhibiting an convex upward turning point and another one at the bottom exhibiting a convex downward turning point. The ensemble of the cords, envelopes and boltropes and the representative curve of the data up to a given moment observed over a sufficiently large interval is referred to as a ‘spatial configuration’.

Qualitative and quantitative indications are obtained from a given spatial configuration by determining which characteristic figures specifically attract and which characteristic figures specifically repel the representative curve of the data, and this is achieved through the examination of numerous and varied past spatial configurations and their subsequent evolutions.

The reasons for which the MLR has been chosen, as mentioned above, are as follows:

• Characteristic figures do not appear within MAs networks;
• Characteristic figures appear clearly within MLRs networks which can be implemented on last-generation PCs;
• MKRs networks, starting with MQRS, are difficult to implement on last-generation PCs, due to limited processing capabilities.

The fact that characteristic figures appear within the network, regardless of the value of the scale parameter, can be exploited to broaden the spectrum of analysis and prediction.

The readability of the graphical display of the network and the representative curve of the data can be improved by using different colors.

Claims

1. A system for the analysis and prediction of the evolution of various types of data such as stock market prices, financial indices and other statistical data, characterized by a dense network of curves constructed mathematically from such data, in which characteristic figures appear.

2. A system according to claim 1, wherein the curves of the network are moving linear regressions or moving regressions other than moving linear regressions.

3. A system according to claim 1, wherein the analysis and prediction of the evolution of the data is achieved through observing the way in which the representative curve of the data is attracted or repelled by the characteristic figures.

4. A system according to claim 1, wherein for the considered data more than one network with different scale parameter values are displayed.

5. A system according to claim 1, wherein multiple colors are used for the display of the network and the representative curve of the data.