Predictive neural network means and method for selecting a portfolio of securities wherein each network has been trained using data relating to a corresponding security

A method for administering an investment portfolio of securities expected to provide, within a predetermined limited period of time, a return superior to performance of a selected market index, the method comprising the steps of:

• estimating an appreciation potential of participating securities of a capital market over the limited period of time using a plurality of predictive neural network means, wherein each of the predictive neural network means estimating appreciation potential of a separate individual security of the participating securities, wherein each of the neural network means has been trained using data relating to a corresponding security so as to adjust weights of each of the neural network means;
• selecting an optimized investment portfolio on the basis of the estimated appreciation potential of the participating securities, determined by the plurality of the neural network means, such that the expected return on the selected investment portfolio over the predetermined period of time is superior to the performance of the selected index;
• and
• receiving information on each security of the selected investment portfolio.

A system comprising one or more computers for selecting a portfolio of securities which is expected to provide, within a predetermined period of time, a return superior to the risk-adjusted performance of a selected market index, comprising:

• a plurality of predictive neural networks for estimating an appreciation potential of participating securities of a capital market over the period of time, wherein each of the neural networks is estimating the appreciation potential of a separate individual security of the participating securities, wherein each of the neural networks has been trained using data relating to a corresponding security so as to adjust weights of each of the neural networks;
• at least one input data preprocessor for generating data for each of the plurality of the neural networks, comprising:
• first means for receiving for each of the participating securities of the capital market current and historical trading information;
• second means for receiving for each of the participating securities of the capital market fundamental information about the security, such fundamental information comprising earnings related data;
• and
• at least one processor for constructing for each of the participating securities of the capital market a set of input data sequences generated from the capital market current and historical trade information and fundamental information, said data sequences being supplied to the the plurality of predictive neural networks for forming an estimate of the appreciation potential of each security over the predetermined period of time, wherein each of the neural networks estimates the appreciation potential of a separate individual security of the participating securities.

A method for computerized selection of a portfolio of securities expected to provide, within a predetermined period of time, a return superior to the risk-adjusted performance of a select market index, comprising the steps of:

• for each participating security of a capital market, electronically acquiring fundamental and price related information;
• electronically analyzing the acquired information for each participating security using a separate predictive neural network which has been trained using data relating to a corresponding security so as to adjust weights of the neural network, to estimate the appreciation potential of each participating security relative to the select market index;
• and
• selecting an investment portfolio on the basis of the estimated appreciation potentials for the participating securities of the capital market such that the expected return on the selected investment portfolio is superior to the performance of the select market index.