2022 meisai topic C idea sharing + translation

background

Market traders frequently buy and sell volatile assets , The goal is to maximize its total return . There is usually a commission for every transaction . Two such assets are gold and bitcoin .

chart 1： The daily price of gold , Dollars per troy ounce . source ： London Bullion Market Association ,2021 year 9 month 11 Japan

chart 2： Bitcoin daily price , US $per bitcoin . source ： NASDAQ ,2021 year 9 month 11 Japan

requirement

A trader asked you to develop a model , The model uses only the daily price stream to date to determine the delivery Should I buy... Every day 、 Holding or selling assets in its portfolio

2016 year 9 month 11 Japan , You will come from 1000 The dollar starts . You will use a five-year trading period , from 2016 year 9 month 11 solstice 2021 year 9 month 10 Japan . On every trading day , The trader will have a portfolio of cash , Gold and bitcoin [C, G, B] In US dollars 、 Troy ounce and bitcoin mean . The initial status is [1000, 0, 0]. Every deal （ Buy or sell ） The commission cost of is the transaction amount . hypothesis αgold = 1% and αbitcoin = 2%. There is no cost of holding assets .

Please note that , Bitcoin can be traded every day , But gold is traded only on open market days , Pricing data files reflect LBMA-GOLD.csv and BCHAIN-MKPRU.csv These two points , Your model should consider this trading schedule .

To develop a model , You can only use the data in the two spreadsheets provided ：LBMA-GOLD.csv and BCHAIN-MKPRU.csv.（ Download it from the official website ）

• Develop a model , Only provide the best daily trading strategy based on price data until that day . stay 2021 year 9 month 10 Use your initial 1000 How much is the value of dollar investment model and strategy ？

• Provide evidence that your model provides the best strategy .

• Determine how sensitive the strategy is to transaction costs . How transaction costs affect strategy and outcomes ？

• At most one memo （ Two pages ） Form your strategy 、 Models and results are communicated to traders

Thought analysis

We can predict the future trend of （ Buy or sell or keep ） These three trading activities make decisions .

First , Forecast the price of gold and bitcoin , You can choose to use a variety of machine learning regression .

here , For example, using a sliding window -xgboost To make predictions , Before use 5 Days of price data to predict 1 Sky price .（ Take the gold price forecast as an example ）

Using tools ：SPSSPRO（ All functions are free 、 Unlimited times and time ）

SPSSPRO- Free professional online data analysis platform

step 1  Upload data

step 2  Null processing

in fact , That is to say, the 10 Data with missing values .

step 3  Process the data , So that the time series can be transformed into pre 5 Days of price data as an independent variable , after 1 Days of price data as a dependent variable .

step 4  Then run xgboost To predict , Incidentally, genetic algorithm is used to estimate the learning rate 、 Regular term parameters for optimization .

Complex model , Long running time , Please be patient ！

result

step 5  After making the forecast , We need to get buy - sell - Maintain this trading strategy , among , Gold only trades on the opening day , This means that on weekends or holidays , The trading status must be holding , The common trading day data of gold and bitcoin can be kept separately for analysis . Suppose gold - Bitcoin is bought and sold together , The main design is a time series data such as a rate of return , such as , We can buy on any day , We can use （ Predict the price of gold one day / The actual price of gold -1） To get a yield , When the increase reaches a certain value , Suggest selling .

Be careful ： The initial state is 【1000,0,0】, And every transaction （ To buy or sell ） The transaction cost of is... Of the transaction amount a%, The gold is 1%, Bitcoin is 2%, that , about 1000 dollar , There are two steps to buying and selling , The actual transaction amount we have is only 940 dollar .

Establish a simple goal plan ： t It is the period from buying to selling

Due to purchase - The sale is ongoing , We need to set up a loop to run .

To achieve a more perfect result , More realistic , Financial risk analysis can be added , similar VaR、CVaR、 Or the use of information entropy , After establishing a perfect investment model , We can use the optimization algorithm to optimize the weight , Such as particle swarm optimization 、 Genetic algorithm (ga) 、 Immune algorithm and so on .