shayandavoodii / onlineportfolioselection.jl

This package offers both traditional benchmark and newly developed Online Portfolio Selection (OPS) algorithms implementation.

Hi there 👋 This is Shayan

Hey, I'm Shayan! I'm super curious and excited about diving into the awesome world of Statistics, AI, and machine learning. I love solving problems, and I see it as a challenge that keeps me going every day.

I've recently completed my M.Sc. in Financial Engineering from the AmirKabir University of Technology (Polytechnic of Tehran). I got into this field because I wanted to mix machine learning with financial investments to make predictions that could beef up investment strategies. After three years of daily learning, I’m still just as stoked as I was when I started this journey!

Presently, my academic interests revolve around the following areas:

• Time Series Analysis
• Time Series Pattern Matching
• Time Series Prediction
• Mathematical Optimization
• Multi-Objective Optimization
• Markov Switching Processes
• Multi-Criteria Decision Making
• Online Portfolio Selection

🎯 Stats

🖇️ Links

onlineportfolioselection.jl's Issues

Fix `ann_sharpe` doc

Needs fix. https://shayandavoodii.github.io/OnlinePortfolioSelection.jl/dev/funcs/#OnlinePortfolioSelection.ann_sharpe-Union{Tuple{T},%20Tuple{T,%20T,%20T}}%20where%20T%3C:Float64

`dricorn` returns an indexing error

This happens when the adequate amount of data is not provided to calculate the market trend. This can be fixed by increasing the number of data samples provided to the agent. However, the error isn't informative enough to know where is the problem and how to fix it.

Precompile using PrecompileTools

Some of the algorithms are needed to be precompiled in order to enhance the FTTX. The potential algorithms to be precompiled are cornu, cornk, and dricorn due to their heavy algorithms.

Remove the `Printf` dependency

The Printf dependency is not necessary and the functionality can be handled using the basic println function. A little bit of extra exprimentation is required to get the formatting correct.

Turn package name on in the make.jl

• Remove the name of the package from the logo and let the documenter.jl put the name below the logo. Currently, the name is being displayed too small.
• Test whether the logo's size can be increased without any displaying issues.

attempt to access 29-element Vector{Float64} at index [0]

Error in running the dricorn function:

julia> res = dricornk(rand(3, 29), rand(29), 5, 10, 5, 10)
ERROR: BoundsError: attempt to access 29-element Vector{Float64} at index [0]
Stacktrace:
 [1] getindex
   @ .\essentials.jl:13 [inlined]
 [2] dricorn_expert(relative_prices::Matrix{…}, asset_ret::Matrix{…}, market_ret::Vector{…}, horizon::Int64, w::Int64, rho::Float64, t::Int64, n_assets::Int64, lambda::Float64)
   @ OnlinePortfolioSelection e:\Julia Forks\OPS.jl\src\Algos\DRICORNK.jl:180
 [3] dricornk(x::Matrix{…}, relpr_market::Vector{…}, horizon::Int64, k::Int64, w::Int64, p::Int64; lambda::Float64, init_budg::Int64, progress::Bool)
   @ OnlinePortfolioSelection e:\Julia Forks\OPS.jl\src\Algos\DRICORNK.jl:97
 [4] dricornk(x::Matrix{Float64}, relpr_market::Vector{Float64}, horizon::Int64, k::Int64, w::Int64, p::Int64)
   @ OnlinePortfolioSelection e:\Julia Forks\OPS.jl\src\Algos\DRICORNK.jl:52
 [5] top-level scope
   @ REPL[53]:1
Some type information was truncated. Use `show(err)` to see complete types.

release

@JuliaRegistrator register

Release notes:

• budgets got removed. It potentially could raise difficulties in understanding the concept (especially in comparison with the sn function.) 4b544b8
• Functions got lowercase.
• @docs utilized for the sake of easier documenting.
• rprt deployment postponed to the next release.
• Unit tests adjusted and useless ones removed.

`sn` have been used wrongly in CAEG.jl

The sn in CAEG.jl has been used wrongly

OnlinePortfolioSelection.jl/src/Algos/CAEG.jl

Line 54 in e656ab2

Sη = sn.(Ref(rel_pr), exp_weights)

. It should be replaced by sn.(exp_weights, Ref(rel_pr)).

Predictions is performed wrongly.

https://github.com/shayandavoodii/OnlinePortfolioSelection.jl/blob/ba5712caa3d4384936d49eb37b2a557e70df2e21/src/Algos/RPRT.jl#LL76C1-L76C1

With assuming t=w=5, relative_prices[:, t-w+1:t-1] is relative_prices[:, 1:4]. Hence, it includes the last 4 RPs instead of 5 (consider that w=5, so 5 last relative prices should be included in the prediction.)
Note that prediction is performed using predict_relative_price(relative_prices[:, t-w+1:t-1]).

Anticor may return weights with very small negative values

Sometimes, anticor may result in very small negative values for some periods during the trading. This only happens when the input prices are so small (which is odd). However, this can be prevented.

Fix the header of `normalizer!` in the documentation.

Header of normalizer! in the documentation doesn't match the definition.

https://github.com/shayandavoodii/OnlinePortfolioSelection.jl/blob/86f47ee720e164f82bef3211bfb20546e86fb9bb/src/Tools/tools.jl#LL81C22-L81C22

Remove duplicated `include`s.

eta should be reordered in the DMR's example

OnlinePortfolioSelection.jl/src/Algos/DMR.jl

Line 97 in 3239215

julia> model = dmr(x, horizon, eta, alpha, n, w);

Remove `t==1` as condition

https://github.com/shayandavoodii/OPS.jl/blob/43ac2e298c212b1d6efc4ad4be7a1494121c44a7/src/Algos/RPRT.jl#L30 t<w is already enough and t==1 is surplus. Because w is supposed to be more than 1 always and it satisfies t==1.

TagBot trigger issue

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

Fix the documentation of FL

A code block is cluttered.

Move `References` of each function to the last.

Currently, the References section of some of the function's documentation is placed before the Example section. Replace these two in any corresponding case.

Documentation got fixed.

Delete the unnecessary @docs at the beginning of the functions' doc.