Data processor for implementing forecasting algorithms
A data processing system and method for developing predictions regarding future asset price movements, based on pattern detection in a time sequence of historical price data. The system includes computer implementation of a kernal regression to effect a smooth estimator of the non-linear price-time relationship. As tested against known patterns, the system provides an expectation regarding a future price movement.
1. A system for processing price data corresponding to a sequence of time for a selected interval and detecting one or more patterns in said price data, said system comprising:
a database communicatively coupled to a data processor and a pattern recognition processor;
said database comprising means for storing said price data in system addressable format;
said data processor responsive to price data stored in said database and comprising means for generating data representing a non-linear function having a smooth, curvilinear characteristic for a range of data within said interval by applying to said price data a smoothing estimator whose parameters comprise a bandwidth, a window length, and a lag time;
said pattern recognition processor in communication with said data processor, operable to receive said data corresponding to said non-linear function, and comprising means for analyzing said smoothed non-linear function to identify one or more patterns of price-time data and generating a results output based on a recognition of said one or more patterns; said pattern recognition processor further comprising means for detecting one or more broadening tops and broadening bottoms, and said results output comprising price movement projections based on said detected one or more broadening tops and broadening bottoms.
2. A system for processing price data corresponding to a sequence of time for a selected interval and detecting one or more patterns in said price data, said system comprising:
a database communicatively coupled to a data processor and a pattern recognition processor:
said database comprising means for storing said price data in system addressable format;
said data processor responsive to price data stored in said database and comprising means for generating data representing a non-linear function having a smooth, curvilinear characteristic for a range of data within said interval by applying to said price data a smoothing estimator whose parameters comprise a bandwidth, a window length, and a lag time;
said pattern recognition processor in communication with said data processor, operable to receive said data corresponding to said non-linear function, and comprising means for analyzing said smoothed non-linear function to identify one or more patterns of price-time data and generating a results output based on a recognition of said one or more patterns; said pattern recognition processor further comprising means for detecting one or more head and shoulders patterns, and said results output comprising price movement projections based on said detected one or more head and shoulders patterns.
3. A system for processing price data corresponding to a sequence of time for a selected interval and detecting one or more patterns in said price data, said system comprising:
a database communicatively coupled to a data processor and a pattern recognition processor:
said database comprising means for storing said price data in system addressable format;
said data processor responsive to price data stored in said database and comprising means for generating data representing a non-linear function having a smooth, curvilinear characteristic for a range of data within said interval by applying to said price data a smoothing estimator whose parameters comprise a bandwidth, a window length, and a lag time;
said pattern recognition processor in communication with said data processor, operable to receive said data corresponding to said non-linear function, and comprising means for analyzing said smoothed non-linear function to identify one or more patterns of price-time data and generating a results output based on a recognition of said one or more patterns; said pattern recognition processor further comprising for detecting one or more triangle tops and triangle bottoms, and said results output comprising price movement projections based on said detected one or more triangle tops and triangle bottoms.
4. A system for processing price data corresponding to a sequence of time for a selected interval and detecting one or more patterns in said price data, said system comprising:
database communicatively coupled to a data processor and a pattern recognition processor:
said database comprising means for storing said price data in system addressable format;
said data processor responsive to price data stored in said database and comprising means for generating data representing a non-linear function having a smooth, curvilinear characteristic for a range of data within said interval by applying to said price data a smoothing estimator whose parameters comprise a bandwidth, a window length, and a lag time;
said pattern recognition processor in communication with said data processor, operable to receive said data corresponding to said non-linear function, and comprising means for analyzing said smoothed non-linear function to identify one or more patterns of price-time data and generating a results output based on a recognition of said one or more patterns; and said pattern recognition processor further comprising means for detecting one or more rectangle tops and rectangle bottoms, and said results output comprising price movement projections based on said detected one or more rectangle tops and rectangle bottoms.
5. A system for processing price data corresponding to a sequence of time for a selected interval and detecting one or more patterns in said price data, said system comprising:
a database communicatively coupled to a data processor and a pattern recognition processor:
said database comprising means for storing said price data in system addressable format;
said data processor responsive to price data stored in said database and comprising means for generating data representing a non-linear function having a smooth, curvilinear characteristic for a range of data within said interval by applying to said price data a smoothing estimator whose parameters comprise a bandwidth, a window length, and a lag time;
said pattern recognition processor in communication with said data processor, operable to receive said data corresponding to said non-linear function, and comprising means for analyzing said smoothed non-linear function to identify one or more patterns of price-time data and generating a results output based on a recognition of said one or more patterns; said pattern recognition processor further comprising means for detecting one or more double tops and double bottoms, and said results output comprising price movement projections based on said detected one or more double tops and double bottoms.