"Enhancing Statistical Significance of Backtests" by Dr. Ernest Chan, Managin...Quantopian
Insufficient historical data is a major hurdle in building a trading model free from data snooping bias. Dr. Chan's talk will discuss several techniques, some borrowed from machine learning, that can alleviate overfitting and enhance the statistical significance of a backtest.
Financial time series forecasting has received
tremendous interest by both the individual and institutional
investors and hence by the researchers. But the high noise and
complexity residing in the financial data makes this job extremely
challenging. Over the years many researchers have used support
vector regression (SVR) quite successfully to conquer this
challenge. As the latent high noise in the data impairs the
performance, reducing the noise could be effective while
constructing the forecasting model. To accomplish this task,
integration of principal component analysis (PCA) and SVR is
proposed in this research work. In the first step, a set of technical
indicators are calculated from the daily transaction data of the
target stock and then PCA is applied to these values aiming to
extract the principle components. After filtering the principal
components, a model is finally constructed to forecast the future
price of the target stocks. The performance of the proposed
approach is evaluated with 16 years’ daily transactional data of
three leading stocks from different sectors listed in Dhaka Stock
Exchange (DSE), Bangladesh. Empirical results show that the
proposed model enhances the performance of the prediction
model and also the short-term prediction gains more accuracy
than long-term prediction.
The objective of this tool was to give a measure of the Value at Risk of the given asset class using techniques like Historical simulation and Monte Carlo simulation. I was involved in the design of a package for estimating the Initial Margin requirement for OTC Derivatives like FX Forward Contracts and Interest Rate Swaps using Historical Value at Risk. I also designed a prototype for running a Monte Carlo simulation on a given stock using Geometric Brownian Motion.
This presentation is for introducing google DeepMind's DeepDPG algorithm to my colleagues.
I tried my best to make it easy to be understood...
Comment is always welcome :)
hiddenmaze91.blogspot.com
Seminar: Gender Board Diversity through Ownership NetworksGRAPE
Seminar on gender diversity spillovers through ownership networks at FAME|GRAPE. Presenting novel research. Studies in economics and management using econometrics methods.
5 Tips for Creating Standard Financial ReportsEasyReports
Well-crafted financial reports serve as vital tools for decision-making and transparency within an organization. By following the undermentioned tips, you can create standardized financial reports that effectively communicate your company's financial health and performance to stakeholders.
how to sell pi coins effectively (from 50 - 100k pi)DOT TECH
Anywhere in the world, including Africa, America, and Europe, you can sell Pi Network Coins online and receive cash through online payment options.
Pi has not yet been launched on any exchange because we are currently using the confined Mainnet. The planned launch date for Pi is June 28, 2026.
Reselling to investors who want to hold until the mainnet launch in 2026 is currently the sole way to sell.
Consequently, right now. All you need to do is select the right pi network provider.
Who is a pi merchant?
An individual who buys coins from miners on the pi network and resells them to investors hoping to hang onto them until the mainnet is launched is known as a pi merchant.
debuts.
I'll provide you the Telegram username
@Pi_vendor_247
BYD SWOT Analysis and In-Depth Insights 2024.pptxmikemetalprod
Indepth analysis of the BYD 2024
BYD (Build Your Dreams) is a Chinese automaker and battery manufacturer that has snowballed over the past two decades to become a significant player in electric vehicles and global clean energy technology.
This SWOT analysis examines BYD's strengths, weaknesses, opportunities, and threats as it competes in the fast-changing automotive and energy storage industries.
Founded in 1995 and headquartered in Shenzhen, BYD started as a battery company before expanding into automobiles in the early 2000s.
Initially manufacturing gasoline-powered vehicles, BYD focused on plug-in hybrid and fully electric vehicles, leveraging its expertise in battery technology.
Today, BYD is the world’s largest electric vehicle manufacturer, delivering over 1.2 million electric cars globally. The company also produces electric buses, trucks, forklifts, and rail transit.
On the energy side, BYD is a major supplier of rechargeable batteries for cell phones, laptops, electric vehicles, and energy storage systems.
2. Elemental Economics - Mineral demand.pdfNeal Brewster
After this second you should be able to: Explain the main determinants of demand for any mineral product, and their relative importance; recognise and explain how demand for any product is likely to change with economic activity; recognise and explain the roles of technology and relative prices in influencing demand; be able to explain the differences between the rates of growth of demand for different products.
The European Unemployment Puzzle: implications from population agingGRAPE
We study the link between the evolving age structure of the working population and unemployment. We build a large new Keynesian OLG model with a realistic age structure, labor market frictions, sticky prices, and aggregate shocks. Once calibrated to the European economy, we quantify the extent to which demographic changes over the last three decades have contributed to the decline of the unemployment rate. Our findings yield important implications for the future evolution of unemployment given the anticipated further aging of the working population in Europe. We also quantify the implications for optimal monetary policy: lowering inflation volatility becomes less costly in terms of GDP and unemployment volatility, which hints that optimal monetary policy may be more hawkish in an aging society. Finally, our results also propose a partial reversal of the European-US unemployment puzzle due to the fact that the share of young workers is expected to remain robust in the US.
how to swap pi coins to foreign currency withdrawable.DOT TECH
As of my last update, Pi is still in the testing phase and is not tradable on any exchanges.
However, Pi Network has announced plans to launch its Testnet and Mainnet in the future, which may include listing Pi on exchanges.
The current method for selling pi coins involves exchanging them with a pi vendor who purchases pi coins for investment reasons.
If you want to sell your pi coins, reach out to a pi vendor and sell them to anyone looking to sell pi coins from any country around the globe.
Below is the contact information for my personal pi vendor.
Telegram: @Pi_vendor_247
Scope Of Macroeconomics introduction and basic theories
RIC-NN: A Robust Transferable Deep Learning Framework for Cross-sectional Investment Strategy
1. The views expressed here are our own and do not necessarily reflect the views of Nomura Asset management.
Any errors and inadequacies are our own.
IEEE Data Science and Advanced Analytics(DSAA) 2020
October 9th, 2020
Masaya Abe1 and Junpei Komiyama2
RIC-NN: A Robust Transferable Deep Learning
Framework for Cross-sectional Investment Strategy
1. Nomura Asset Management Co., Ltd.
2. New York University
Kei Nakagawa1,
2. 1. Introduction and Motivation
2. Data and Methodology
3. Experimental Results
4. Conclusion
Agenda
1
4. 3
Motivation – Stock Return Prediction
・ Those studies are mostly based on time series prediction and not
practical.
Therefore, we propose a practical framework for stock return
forecasting by cross-sectional prediction with deep learning.
・ Stock returns follow “random walk”, and it is difficult to predict
them with time series prediction for a long time.
・ Many study of stock return prediction using machine learning.
5. 4
Motivation – Stock Return Prediction
クロスセクションで予測
Stock A Stock B ・・・
T+1
T
T-1
・
・
・
Buyastockwithabsolutely
highreturn.
Buy stocks with relatively high scores
based on certain criteria
Investment Universe
TimeSeries
Cross SectionOrthogonal
1≤ 𝑛 ≤N
6. 5
・ Over 300 factors until 2012 [Harvey et al. 2017]
・ In practice, we predict future relative stock returns (scores) by
combining various factors.
Cross-sectional Prediction
・ Criteria for describing the score is called Factor in Finance
Cross-sectional Investment Strategy
7. ROE
1 Month Return
●
●
●
Score
Value
Growth
Quality
Momentum
・Linear Regression
Factor
Candidates
Factor Classification
by human
Calculate
Relative Goodness
Cross-sectional Investment Strategy
Relative Stock Returns
・Rank IC (Spearman correlation):
6
・Cross-sectional Investment Strategy in practice
8. 7
Cross-sectional Investment Strategy
(2) A few studies use deep learning for cross sectional investment
strategy but they suffer from overfitting.
・ Challenges in cross-sectional investment strategy
(1) Traditionally, assume the relationship between factors and
returns is linear but actually, it's non-linear. Levin[1996]
(3) Not enough data for deep learning can be gathered in some markets.
9. RIC-NN: Our methodology
Score
Input Output
(3) Deep Transfer Learning
Loss
rank IC
Stop
Epoch
Time step t-1 Time step t
𝒗𝒊 𝒗 𝒇
(2) Weight Initialization and Stopping
Stop
𝒗 𝒇
Loss
(1) Multi-factor Deep Learning Approach
Score
Stock
Input Output
Score
Stock
Input Output
Source Domain
Target Domain
Transfer
Stock Factor
Stock
Factor
Stock
Stock
Factor
Initialization
𝒗𝒊
NorthAmericaAsiaPacific
Initialization
8
11. ・ DL for stock return prediction easily overfits to training data.
✓ Use early stopping to control the fitness to the past data.
RIC-NN: Weight Initialization and Stopping
・Use RankIC (Spearman
correlation) in terms of the fitness.
-> intuitive and controllable.
Loss
rank IC
Epoch
Time step T-1 Time step T
Stop
Loss
Initialization
・ Epoch-based stopping,
Risk of overfitting or underfitting
because the training speed varies.0.20.16
Stopping: the rank IC reaches 0.20.
Initialization: Use the model of timestep t-1 when rank IC is 0.16.
c.f.: fitness of a good portfolio to
future return is around 0.10
Initialization
Our Proposed (RIC-NN)
✓ Training at time step t:
10
12. ScoreRIC-NN
Factor
Stock
Input Output
Score
Factor
Stock
Input Output
North America Stock Market Asia Pacific Stock Market
(Source Domain) (Target Domain)
Transfer
RIC-NN
・Augment the model using the knowledge of a larger market.
✓ Use transfer learning
RIC-NN: Deep Transfer Learning
We want to capture the asymmetric structure between the two markets.
11
14. ・ We use the 20 factors that are often used in practice.
✓ Calculated for each regional index constituents
- MSCI North America Index (NA)
- MSCI Pacific Index (PA)
Features (Various Factors)
No. Feature (Factor) No. Feature (Factor) No. Feature (Factor)
1 Book-to-market Ratio 8 Return on Invested Capital 15 EPS Revision(1 month)
2 Earnings-to-price Ratio 9 Accruals 16 EPS Revision(3 months)
3 Dividend Yield 10 Total Asset Growth Rate 17 Past Stock Return(1 month)
4 Sales-to-price Ratio 11 Current Ratio 18 Past Stock Return(12 months)
5 Cash flow-to-price Ratio 12 Equity Ratio 19 Volatility
6 Return on Equity 13 Total Asset Turnover Rate 20 Skewness
7 Return on Asset 14 CAPEX Growth Rate
※ Monthly data
Data sources: Namely, Compustat, WorldScope, Thomson Reuters, I/B/E/S and EXSHARE.
13
15. Features (Various Factors)
・ Cumulative returns in NA(left-side) and PA(right-side) on 20 factors
・ Return of each factor varies largely over time.
14
16. Problem Formulation
MSE 𝑡 =
1
𝐾
𝑡′=𝑡−𝑁
𝑡−1
𝑖∈𝑈 𝑡′
𝑟𝑖,𝑡′+1 − 𝑓 𝒗𝑖,𝑇; 𝜽 𝑇+1
2
𝑁 = 120 (10 years)
𝐾 =
𝑡′=𝑡−𝑁
𝑡−1
𝑈 𝑡′
・ We define the problem as a regression problem to minimize MSE.
✓ Approximate function 𝑓 ∙ with the parameter 𝜽 𝑇+1 that maps 𝒗𝑖,𝑇 to 𝑟𝑖,𝑇+1
𝑓 𝒗𝑖,𝑇; 𝜽 𝑇+1 → 𝑟𝑖,𝑇+1
✓ Train the models using the data of the latest 120 time steps from the
past 10 years.
15
𝒗𝑖,𝑇 :Augmented factors
𝜽 𝑇+1:NN Weights
𝑟𝑖,𝑇+1:Relative Stock return
17. ✓ 20 factors:
Problem Formulation
・ Given a stock 𝑖 at month 𝑇 (𝑖 ∈ 𝑈 𝑇: a regional index constituents at 𝑇)
𝒙𝑖,𝑇 ∈ 𝑅20
✓ Features: 20 factors and preprocessed factors
𝑥/ 𝑅 𝑦 ≔ 2(𝑥 − 𝑦)/( 𝑥 + 𝑦 )
✓ Output variable: scaled one-month-ahead stock return (Score)
Scale to the range [0,1]
Pre-processing & Feature augmentation
𝒗𝑖,𝑇 = (𝒙𝑖,𝑇, 𝒙𝑖,𝑇−3, … , 𝒙𝑖,𝑇−12, 𝒙𝑖,𝑇/ 𝑅
𝒙𝑖,𝑇−3, … , 𝒙𝑖,𝑇/ 𝑅
𝒙𝑖,𝑇−12) ∈ [0,1] 𝟏𝟖𝟎
𝑟𝑖,𝑇+1 ∈ [0,1]
16
・Most of factors are updated quarterly ・Time difference between the present and each quarter ago
18. ・ Architecture of RIC-NN is quite standard
✓ Fully-connected feedforward neural networks
✓ 6 Hidden layers: { 150 – 150 – 100 – 100 – 50 – 50 }
Dropout rates: (50% – 50% – 30% – 30% – 10% – 10%)
✓ Activation function: ReLU function
✓ RIC-NN(Transfer Learning: TF)
Compared Models
We use the weights of the first four layers that are trained
in the source region as the initial weight of the target region.
Our Proposed (RIC-NN)
17
20. 19
Prediction Period
✓ 14 years (from January 2005 to December 2018)
✓ Updated by sliding one-month-ahead and performed a monthly forecast.
14 years ( 168 months )
・・・
October 2018
:
November 2008
December 2018
November 2018
𝜽 𝑇
∗𝒗𝑖,𝑇
December 2004
:
January 1995
February 2005
January 2005
𝒗𝑖,𝑇 𝜽 𝑇
∗
November 2004
:
December 1994
January 2005
December 2004
𝜽 𝑇
∗
Scores
Training
120 set
𝑓 𝒗𝑖,𝑇; 𝜽 𝑇
∗
argmin
𝜽
MSE 𝑇
Features: 𝒗𝑖,𝑇
𝒗𝑖,𝑇
22. ・ Simple portfolio strategies
Performance Measure
✓ Long Portfolio Strategy
✓ We make quintile portfolios.
- Buy (Long) the top 1/5 score stocks with equal weighting
- Benchmark: the average return of all stocks
→ Relative performance evaluation
1 2 3 4 5
Relative goodness
Investment Universe
21
✓ Long Short Portfolio Strategy
- Buy the top 1/5 score stocks with equal weighting
- Sell the bottom 1/5 score stocks with equal weighting
→ Absolute performance evaluation (No benchmark)
23. Performance Measure
・ We use the following measures for Long (resp. Long-Short) portfolio.
✓ Alpha Return ≔ ς 𝑡=1
𝑇
1 + 𝑟𝑡
12/𝑇 − 1
✓ 𝑇𝐸 Risk ≔
12
𝑇−1
σ 𝑡=1
𝑇
𝑟𝑡 − 𝜇 𝑟
2
✓ 𝐼𝑅 R/R ≔ Alpha(Return)/𝑇𝐸(Risk)
portfolio return – benchmark return
𝜇 𝑟: Average of 𝑟𝑡
𝑟𝑡:
✓ MaxDD ≔ min
𝑘∈[1,𝑇]
(0,
𝑊𝑘
𝑃𝑜𝑟𝑡
max
𝑗∈ 1,𝑘
𝑊𝑗
𝑃𝑜𝑟𝑡 − 1)
𝑊𝑘
𝑃𝑜𝑟𝑡
: Cumulative return of the portfolio
22
24. Linear Linear
LR RF GB
NN
(Epoch)
RIC-NN
RIC-NN
(TF from PF)
LR RF GB
NN
(Epoch)
RIC-NN
RIC-NN
(TF from NA)
Alpha [%] 0.62 0.79 0.87 0.82 1.23 1.20 5.35 3.79 5.59 4.34 5.25 5.78
TE [%] 5.40 5.14 4.36 4.48 4.14 4.43 5.17 5.75 5.00 4.18 4.20 3.95
IR 0.11 0.15 0.20 0.18 0.30 0.27 1.04 0.66 1.12 1.04 1.25 1.46
MaxDD [%] -21.84 -24.57 -18.10 -17.41 -14.37 -20.57 -11.53 -11.43 -8.86 -9.37 -7.51 -3.37
Linear Linear
LR RF GB
NN
(Epoch)
RIC-NN
RIC-NN
(TF from PF)
LR RF GB
NN
(Epoch)
RIC-NN
RIC-NN
(TF from NA)
Return [%] 2.24 1.71 2.29 2.10 3.86 2.16 10.27 7.78 10.79 8.52 9.81 10.95
Risk [%] 10.90 11.51 8.62 9.47 7.85 9.52 9.23 9.65 8.68 7.78 7.83 7.14
R/R 0.21 0.15 0.27 0.22 0.49 0.23 1.11 0.81 1.24 1.10 1.25 1.53
MaxDD [%] -34.73 -42.21 -34.35 -34.49 -21.26 -39.35 -18.07 -18.66 -13.92 -19.74 -11.06 -8.89
Long
MSCI North America MSCI Pacific
Nonlinear Nonlinear
Long-Short
MSCI North America MSCI Pacific
Nonlinear Nonlinear
Experimental Results (1/3)
・ NA: RIC-NN without transfer learning performed best.
・ PA: RIC-NN with transfer learning performed best.
・ NA as a source domain enhances the performance of PA,
not vice versa.
23
25. Experimental Results (2/3)
・ While NN at epoch 50 performs better in NA, NN at epoch 60 performs better in PA.
・ NN(Epoch) is very sensitive to the choice of the epoch.
・ RIC-NN outperforms epoch-based stopping: rank IC controls the fitness of
the stock prediction models consistently.
*These epochs are chosen so that the rank IC reaches 0.20 during the training of the first time step.
24
40 50 56 60 80 40 46 50 60 80
Alpha [%] 1.23 0.18 1.48 0.82 1.25 0.70 5.25 4.13 4.34 4.28 4.52 2.99
TE [%] 4.14 4.52 4.35 4.48 4.49 4.14 4.20 4.36 4.18 4.73 4.34 4.06
IR 0.30 0.04 0.34 0.18 0.28 0.17 1.25 0.95 1.04 0.90 1.04 0.74
MaxDD [%] -14.37 -22.67 -13.48 -17.41 -20.98 -15.94 -7.51 -8.08 -9.37 -7.16 -7.45 -7.52
40 50 56 60 80 40 46 50 60 80
Return 3.86 0.67 3.24 2.10 2.02 3.10 9.81 8.89 8.52 8.97 9.78 6.15
Risk 7.85 9.08 10.06 9.47 9.05 7.73 7.83 7.63 7.78 8.05 7.73 7.18
Return/Risk 0.49 0.07 0.32 0.22 0.22 0.40 1.25 1.16 1.10 1.11 1.26 0.86
MaxDD -21.26 -40.09 -26.20 -34.49 -31.62 -23.47 -11.06 -13.07 -19.74 -12.17 -14.70 -13.44
MSCI Pacific
MSCI Pacific
NN(Epoch)Long
Long-Short
MSCI North America
MSCI North America
RIC-NN
NN(Epoch)
RIC-NN
RIC-NN
NN(Epoch)
RIC-NN
NN(Epoch)
*
*
*
*
26. Experimental Results (3/3)
25
・ We select these funds by querying Bloomberg fund
screening search with the following conditions:
・We compare the performance of RIC-NN with major funds where
the investments involve decision-making by human experts.
・ We select the top 5 funds in terms of the total assets and
calculate average total return series of these funds, including the
trust fees.
Fund Asset Class Focus: Equity Asset Class
Fund Geographical Focus: North America Region (resp. Asian Pacific Region)
Fund Type: Open-End-Funds
Currency Base: US dollar
Market Cap Focus: Large-cap and Mid-cap focus
27. 26
5 Funds
(average)
RIC-NN
RIC-NN
(After Cost
Reduction)
5 Funds
(average)
RIC-NN
RIC-NN
(After Cost
Reduction)
Return [%] 5.90 9.09 7.79 7.88 12.08 9.44
Risk [%] 14.91 17.78 17.78 17.58 17.23 17.23
R/R 0.40 0.51 0.44 0.45 0.70 0.55
MSCI Pacific
Long
MSCI North America
Experimental Results (3/3)
・RIC-NN after transaction costs outperformed the top 5 funds average.
28. 27
・ Machine learning methods in finance
✓ Survey: Bahrammirzaee (2010), Cavalcante et al. (2016)
→ Most of all studies are time series prediction
・ Neural Networks for cross sectional investment strategy
✓ Classical Neural Networks(Epoch): Levin (1996)
✓ Deep Neural Networks(Epoch): Abe and Nakayama (2018),
Nakagawa et al (2018, 2019)
Related Work
However, as we confirmed epoch based networks are very sensitive
to the choice of the epoch.
29. ・ We have proposed a new cross sectional stock return prediction framework
called RIC-NN.
by introducing three practical ideas:
(1) A nonlinear multi-factor approach is better than a linear approach.
(2) Rank IC-based stopping outperforms epoch-based stopping.
(3) Multi-region transfer learning works well.
・ Better return of the portfolio, better control of the fitness of the model to
the past dataset.
Conclusion
28