Yu Peng Email & Phone Number

New York City, New York, Us

New York City, New York, Us

New York City, New York, Us

Hoboken, Nj, Us

Constrained portfolio optimization with Recurrent Reinforcement Learning and Particle Swarm Optimization•Developed an adaptive RRL-PSO portfolio rebalancing decision system with a market condition stop-loss retraining mechanism. Used Calmar ratio and log-return alternately or the combination as the objective function in training process. System will switch objective function smartly depending on profit driven or risk management driven strategies, and these strategies was determined by volatility forecast(EGarch-Ann-NLP).•This trading system is capable of generating both long and short signals and handling the common portfolio constraints. It was tested on highly-liquid ETFs.•Result showed that the proposed portfolio trading system outperforms the benchmarks consistently especially under high transaction cost conditions.•Improved algorithm from immediate reward feedback to multiple reward feedbacks, and currently working on the improvement of convergence.

Hoboken, Nj, Us

Dark Pool imbalanced orders Classification• Classify dark pool imbalanced orders(10:1) in order to minimize the transaction cost • Implemented Adaboost technique to change underlying data distribution and classifiy in the re-weighted data space iteratively, in order to boost weak order classifiers into strong ones.• Improved Adaboost into MEBoost technique, which mixed decision tree and extra tree classification learners alternatively on the training data set with boosting weights.•This algorithm increased recall and precision rate by roughly 20 percent

Hoboken, Nj, Us

State College, Pa, Us

Us