<![CDATA[Stock investment is one of the most populer investments nowadays. This kind of investment has the "high risk high return" characteristic which come up with a threat of loss for stock investors. There are lots of paper have been implemented related to the estimation of stock price movements, but researchers focus more on technical analysis rather than fundamental analysis which is no less essential. One of the populer methods with a fundamental approach is Price Earning Ratio (PER) method. Extreme Learning Machine is a proven method of forecasting stocks with high performance and relatively low learning speed, but this method has weaknesses in determining random weights and biases that can reduce its stability. Kernel Extreme Learning Machine offers the utilization of kernel functions that can provide high stability and performance, but with relatively low learning speed. The results of this paper provide the optimal Mean Absolute Precentage Error (MAPE) is 2.78021%, with 8 features, training and testing data ratio 90%: 10%, using the Polynomial kernel function with a value of parameter 1, and using a regularization coefficient (λ) 1000. Nested Cross Validation evaluation was also performed which provide the MAPE value is 6.385713%.]]>
Copyrights © 2020
