“Left Shoulder” Detection in Korea Composite Stock Price Index Using an Auto-Associative Neural Network

We propose a neural network based “left shoulder” detector. The auto-associative neural network was trained with the “lelt sliouldcr” patterns obtained from the Korea Composite Stock Price Index, and then tested out-of- sample with a reasonably good result. A hypothetical investment strategy besed on the detector achieved a return of 124% in comparison with 39% return from a buy and hold strategy.