Feature Learning for Stock Price Prediction Shows a Significant Role of Analyst Rating

Efficient Market Hypothesis states that stock prices are a reflection of all the information present in world and generating excess returns is not possible by merely analysing trade data which already available to public. Yet further research rejecting this idea, rigorous literature review was conducted set five technical indicators 23 fundamental identified establish possibility on market. Leveraging these points various classification machine learning models, trading 505 equities US S&P500 over past 20 years analysed develop classifier effective for our cause. From any given day, we were able predict direction change price 1% up 10 days future. The predictions had an overall accuracy 83.62% with precision 85% buy signals recall 100% sell signals. Moreover, grouped their sector repeated experiment see if grouping similar assets together positively effected results but concluded it showed no significant improvements performance—rejecting idea sector-based analysis. Also, using feature ranking could identify even smaller 6 while maintaining accuracies as from original 28 features also uncovered importance buy, hold analyst ratings they came out be top contributors model. Finally, evaluate effectiveness real-life situations, backtested FAANG (Facebook, Amazon, Apple, Netflix & Google) modest strategy where generated high above 60% term testing dataset. In conclusion, proposed methodology combination purposefully picked shows improvement previous studies, model predicts changes 10th day confidence enough buffer build robotic system.