I need u to help me out for a machine Learning project on R :
Will send u the class material, there are all the codes needed. We just have to change a bit and make some relevant work.
Dynamic portfolio backtesting project:
In this case, the focus is more on the portfolio strategy and the dynamic backtesting protocol. Several routes are possible:
• either test many simple strategies and compute a deflated Sharpe ratio test; • or propose ML-based strategies.
Any static allocation (i.e. that does not change in time) will fail the project.
Allocation process :
Please be specific on the way the portfolio is constructed. Make sure the following elements are clear:
1. the signal on which the strategy is built (ML-based (hopefully), or other);
2. the way assets are integrated in or barred from the portfolio (asset selection);
3. the way the assets are weighted inside the portfolio (do you use optimisation, and if yes, which one precisely - include the constraints if there are any).
Performance metrics :
The possible output usually has the following form:
• the usual: average returns, volatility, Sharpe ratio, Value-at-Risk, turnover;
• turnover-adjusted Sharpe ratio (i.e., transaction cost-adjusted);
• deflated Sharpe ratios if you test a lot of strategies;
• maximum drawdown and MAR ratio (possibly transaction cost-adjusted as well);
• other metrics are possible (e.g., alphas in CAPM or Fama-French regressions; semi-volatility; or others)
Any combination of the above is obviously welcome. Plots of the portfolio values are always appreciated (via the cumprod() function) - cumulative returns are accepted as well. Also, choose one unique frequency to compute these indicators. Annualised values are a good choice. Monthly values are also valid.