Pesquisa sobre big data e machine learning ao trading

The user has requested enhancement of the downloaded file. Big Data and Machine Learning Models in Quantitative Finance 17th Asia-Pacific Conference on Global Business, Economics, Finance & Social Sciences AP18Hong Kong Conference in Hong Kong, SAR. January 19-20, 2018 Dr. Carlin Chu The Open University of Hong Kong ccfchu@ouhk. edu. With data volumes and varieties constantly increasing, especially from social media and the Internet of Things (IoT), that's a key consideration. Computing power. Hadoop's distributed computing model processes big data fast. The more computing nodes you use, the more processing power you have. Fault tolerance. You can easily grow your system to handle more data simply by adding nodes. Little administration is required. Access Method NoSQL (Not Only SQL database) 13 D) Commonly used Machine Learning techniques 14 Coverage of ML models 15 Model complexity vs Accuracy Out-of-Sample Error 16 E) Review Exercise Data Structured/ Unstructured Individual/ Business generated/ Sensor ML Techniques Picture/Video U I Blog/Twitter/News U I/B Email / Mobile data text message U I Text Mining E-commerce transactions (Purchase patterns) S B Market basket analysis Co-occurrence Web search trend (Prob of getting more popular) S I Logistic Regression Trading data (Up/Down trend) S B Classification Macroeconomic data (Next quarter GDP) S B Regression Foot-fall data/People count (number in next period) S S Time series analysis Neural Network / Convolution network Sentiment analysis 17 A) Characteristics of Big Data Content B) Data for quantitative finance C) Reasons for Machine Learning (ML) D) Commonly used ML techniques E) Review exercise (Structured/unstructured; Individual/Business/Sensor, Analysis Methods) F) Main steps in building ML models G) Example of using alternative data/ non-traditional data for trading ? H) Some misconceptions perceived by the public towards Big data & ML in Finance (quoted by JP Morgan) I) Current issues addressed in ML Quantitative Finance 18 F) Main steps in building Big Data ML Trading models algorithm 19 Case Study 1: Twitter Sentiment usage • iSentium – real-time sentiment, Twitter messages.

– sentiment search engine, judge the potential market impact of a tweet, a news article, or other social media activities. • J. Select to the most representative 100 stocks of S&P 500, filtered using tweet volume and realized volatility measures. Tweets are assigned a sentiment score using a patented NLP algorithm. Aggregating tweet scores, a sentiment level is produced per minute between 8:30 AM and 4:30 PM every day. Sentiment for the day is aggregated using an exponentially weighted moving average over the past ten days. S&P 500 returns are forecasted using a linear regression over the sentiment scores for the past two days. g. percentage change, or normalize data for an average spend). • There is also meaningful seasonality in the data. For instance spend data is slightly lower on weekends and is higher during month of November.

After winsorizing to 5th-95th percentile …. We keep the stock in the basket till 5 days after the announcement date. Stocks are equal weighted. Good Bad 31 Case study 4: Satellite Imagery of Parking Lots and Trading Retail Stocks (RS Metrics) • Analyzes geospatial data drawn from various sources including satellites, drones and airplanes. • uses ~10 high-resolution satellites, orbiting the Earth and imaging between 11 AM and 1:30 PM. • Estimate retail traffic (e. standard deviations, a sell signal is issued. • Trading strategy for each stock was as follows: – a position starts from the date on which a signal (1, 0, -1) is produced by RS Metrics and the position is held until a change in the signal. – A signal of 1 triggers a long stock and short US Retail index benchmark position; – -1 triggers a short stock and long US Retail index position.

– For a 0 signal, we are invested in cash (close the position). • Aggregated performance since 2013, the strategy delivered a Sharpe ratio of 0. g. Tweets, social messages, blog message) ? No difficult to ‘interpret’ large piece of data 4. Machine Learning algorithms are always black box ? No. Interpretable (Tree); Semi-interpretable(SVM); Black box (Neural Network/Deep network) Difficult to fully explained the impact from every single parameters, but outcome/result is explainable by the structure as a whole. I) Current issues - ML in Quantitative Finance • Understandability of the derived logics – Deep learning, Blackboxed model • Any governance in using the models in the industry ? Potential of Huge loss • Withstand to extreme/abnormal market situation? tail risk analysis • Effective inclusion of unstructured data (e. html • • Machine Learning vs. Quants: The Advantages of Machine Learning in Finance, MARCH 13, 2016 http://www.

bradfordcross. com/blog/2016/3/13/machine-learning-vs-quants-the-advantages-ofmachine-learning-in-finance-1 • • How is machine learning used in quantitative finance? -- Quora https://www. quora.