Prediction in Stock Trading

Slides:



Advertisements
Similar presentations
Experiments with MATLAB Experiments with MATLAB Google PageRank Roger Jang ( 張智星 ) CSIE Dept, National Taiwan University, Taiwan
Advertisements

A Heuristic Bidding Strategy for Multiple Heterogeneous Auctions Patricia Anthony & Nicholas R. Jennings Dept. of Electronics and Computer Science University.
Task Assignment and Transaction Clustering Heuristics.
Models Physical: Scale, Analog Symbolic: Drawings Computer Programs Mathematical: Analytical (Deduction) Experimental (Induction)
Information Analysis. Introduction How can we judge the information content of an idea? More concretely: –Is this a signal we can profitably deploy.
Resource Allocation Problem Reporter: Wang Ching Yu Date: 2005/04/07.
Statistical Arbitrage Ying Chen, Leonardo Bachega Yandong Guo, Xing Liu March, 2010.
ENGR-25_HW-01_Solution.ppt 1 Bruce Mayer, PE Engineering/Math/Physics 25: Computational Methods Bruce Mayer, PE Licensed Electrical.
Highline Class, BI 348 Basic Business Analytics using Excel, Chapter 01 Intro to Business Analytics BI 348, Chapter 01.
Decision Trees Jyh-Shing Roger Jang ( 張智星 ) CSIE Dept, National Taiwan University.
Computational Complexity Jang, HaYoung BioIntelligence Lab.
Haksun Li
Mean Reverting Asset Trading Project Presentation CSCI-5551 Grant Meyers.
Soft Computing methods for High frequency tradin.
Maximum Likelihood Estimate Jyh-Shing Roger Jang ( 張智星 ) CSIE Dept, National Taiwan University.
Discussions on Audio Melody Extraction (AME) J.-S. Roger Jang ( 張智星 ) MIR Lab, CSIE Dept. National Taiwan University.
Simulation of Stock Trading J.-S. Roger Jang ( 張智星 ) MIR Lab, CSIE Dept. National Taiwan University.
Linear Classifiers (LC) J.-S. Roger Jang ( 張智星 ) MIR Lab, CSIE Dept. National Taiwan University.
Trading Simulator Dry Run.
Commonly Used Datasets for ML
Search in Google's N-grams
Maximum Likelihood Estimate
Quadratic Classifiers (QC)
DP for Optimum Strategies in Games
Date of download: 10/13/2017 Copyright © ASME. All rights reserved.
張智星 (Roger Jang) 台大資工系 多媒體檢索實驗室
Query by Singing/Humming via Dynamic Programming
National Taiwan University
Intro to Machine Learning
Buying Fashion Fashion Marketing.
Particle Swarm Optimization
National Taiwan University
Feature Selection for Pattern Recognition
Quantitative Methods Dr. Aravind Banakar –
Quantitative Methods Dr. Aravind Banakar –
Quantitative Methods
Quantitative Methods
Quantitative Methods
Quantitative Methods
Quantitative Methods
Quantitative Methods
The Modeling Process Objective Hierarchies Variables and Attributes
Disjoint Subsets Investing Efficiency Analysis
XIV Investment Strategy Optimization
Search in OOXX Games J.-S. Roger Jang (張智星) MIR Lab, CSIE Dept.
Applications of Stacks and Queues for Constraint Satisfaction Problems
السيولة والربحية أدوات الرقابة المالية الوظيفة المالية
Buying Fashion Fashion Marketing.
Mean Reverting Asset Trading
Mean Reverting Asset Trading
Intro to Machine Learning
Deep Neural Networks (DNN)
Hierarchical Clustering
Machine Learning in FinTech
National Taiwan University
Downhill Simplex Search (Nelder-Mead Method)
Predictive Keyword Scores to Optimize Online Advertising Campaigns
Applications of Heaps J.-S. Roger Jang (張智星) MIR Lab, CSIE Dept.
Query by Singing/Humming via Dynamic Programming
Insertion Sort Jyh-Shing Roger Jang (張智星)
Scientific Computing: Closing 科學計算:結語
5/2/14 “Market Cap” What is a company’s market cap? (Definition in your own words.) What is the company with the highest market cap in the following industries.
Performance, Applications, Security
Naive Bayes Classifiers (NBC)
Game Trees and Minimax Algorithm
Introduction to algo quant, an integrated trading research tool
National Taiwan University
Sorting Algorithms Jyh-Shing Roger Jang (張智星)
Storing Game Entries in an Array
Pre and Post-Processing for Pitch Tracking
Presentation transcript:

Prediction in Stock Trading J.-S. Roger Jang (張智星) MIR Lab, CSIE Dept. National Taiwan University jang@mirlab.org, http://mirlab.org/jang 2019/5/3

Types of Predictions in Stock Trading Up/down prediction: Classification Use historical data as groundtruth (GT) Price prediction: Regression Use historical data as GT Timing prediction for “buy” and “sell” to maximize profit Set up groundtruth  How? Use historical data to do prediction Historical data from 技術面、消息面、基本面… Backtest (回測)

GT for “Buy” and “Sell”: Heuristic Approaches Heuristic approaches to set up GT 慣量型 連漲三天 買!(因為還會上漲…) 連跌三天 賣!(因為還會下跌…) 逆勢型 連漲五天  賣!(因為快開始跌了…) 連跌五天  買!(因為快開始漲了…) 謹慎小心型 上漲超過買價的10%  賣! 下跌超過賣價的10%  買! Performance can be evaluated by backtest. Quiz!

GT for “Buy” and “Sell”: Optimal Approach Problem definition Given the price curve over a long period of time, find “buy” and “sell” points (where you can look into the future) such that the overall profit is maximized. Assumptions 每一次都全買或全賣 根據開盤價格來買或賣,一定會成交 必須計算交易費 Approaches Optimal search: Dynamic programming Heuristic search: Genetic algorithms, simulated annealing, randomized search?

Set Up GT for Max. Profit in HW05 HW05: How can we set up GT for “buy” and “sell” points, such that the overall profit can be maximized? MATLAB functions Function submitted by you: actionVec = myOptimAction(priceVec, transFeeRate); Evaluation by our judge system returnRate = profitEstimateOpen(priceVec, transFeeRate, actionVec); Scoring is based on your ranking! All price data is here!

Prediction of GT Derived by Optimal Methods 2019/5/3 For a given date Predict all the future price using historical data Run DP to predict all the “buy” and “sell” points based on the predicted price Execute the predicted plan for tomorrow Future price prediction Time series prediction (regression) One-step-ahead prediction 𝑥 𝑛 =𝑓( 𝑥 𝑛−𝑑 , 𝑥 𝑛−𝑑−1 ,…, 𝑥 𝑛−1 ), 𝑥 𝑛 ∈𝑅. Accumulated prediction: 𝑥 1 , 𝑥 2 , 𝑥 3 , 𝑥 4 , 𝑥 5 , 𝑥 6 , 𝑥 7 , 𝑥 8 , 𝑥 9 , 𝑥 10 K-step-ahead prediction

Prediction of GT Derived by Heuristics 2019/5/3 For a given date Predict “up for m days” or “down for n days” Execute the predicted plan for tomorrow if the prediction tell us to “buy” or “sell” Future price prediction Prediction of up/down (classification) One-step-ahead prediction: 𝑥 𝑛 =𝑓 𝑥 𝑛−𝑑 , 𝑥 𝑛−𝑑−1 ,…, 𝑥 𝑛−1 , 𝑥 𝑛 ∈ ±1 . Accumulated prediction: 𝑥 1 , 𝑥 2 , 𝑥 3 , 𝑥 4 , 𝑥 5 , 𝑥 6 , 𝑥 7 , 𝑥 8 , 𝑥 9 , 𝑥 10 K-step-ahead prediction

References and Resources 2019/5/3 MATLAB Time Series Forecasting Using Deep Learning Time Series Prediction and Forecasting for Prognosis Time Series Analysis … Python 7 methods to perform Time Series forecasting A comprehensive beginner’s guide to create a Time Series Forecast 11 Classical Time Series Forecasting Methods in Python