joelowj/machine-learning-and-reinforcement-learning-in-finance

Machine Learning and Reinforcement Learning in Finance New York University Tandon School of Engineering

Machine-Learning-and-Reinforcement-Learning-in-Finance

Guided Tour of Machine Learning in Finance

1. Euclidean Distance Calculation
2. Linear Regression
3. Tobit Regression
4. Bank defaults prediction using FDIC dataset

Fundamentals of Machine Learning in Finance

1. Random Forests And Decision Trees
2. Eigen Portfolio construction via PCA
3. Data Visualization with t-SNE
4. Absorption Ratio via PCA

Reinforcement Learning in Finance

1. Discrete-time Black Scholes model
2. QLBS Model Implementation
3. Fitted Q-Iteration
4. IRL Market Model Calibration

Overview of Advanced Methods of Reinforcement Learning in Finance

machine-learning-and-reinforcement-learning-in-finance's Issues

Fitted Q-Iteration (80/100)

Current submission

Reward Function 20/20
Replicating Portfolio 20/20
Coefficient Matrix S 20/20
Coefficients Vector M 20/20
Fitted Q-Iteration Option Price 0/20

I haven't had time to figure out the issue with Fitted Q-Iteration Option Price but this would suffice for those who need to pass this assignment.

Tobit Regression Assignment (80/100)

Current submissions

Tobit Model Weights 20/20
Tobit Model: loss, standard error, L1 error . 20/20
Linear Regression with non-linear data 20/20
Linear Regression Model Score 20/20
Neural Network with Non-Linear Data 20/20

My implementation for Neural Network with Non-Linear Data did not clear the grader.
Feel free to start a discussion if you managed to clear the grader or have an even better implementation.

Course 3 and Course 4

When are the assignments of the remaining courses (course 3 and 4) going to be uploaded?

Eigen Portfolio construction via PCA (75/100)

Current submission

Asset Returns Calculation 15/15
Calculate Variances of Returns 15/15
Eigen Portfolio Weights 15/15
Second Eigen Portfolio Weights 15/15
Eigen Portfolios: Sorted by Sharpe Ratio 25/25
Portfolio with the Highest Sharpe Ratio 15/15

Having issue clearing the grader for sorted by Sharpe ratio.
The sorted sharpe ratio portfolio I am getting is 42, 104, 2, 94, 9, ... We know that 42 is definitely right, since 42 cleared the section of Portfolio with the Highest Sharpe Ratio.

DP QLBS not work when T is in a larger scale

When I set T=252 which is a usually chosen transaction day, the result will lead to a wrong result of more than 1000.

Absorption Ratio via PCA (75/100)

Current submission

Exponentially Weighted Returns 25/25
Linear Auto Encoder 25/25
Portfolio Weights 25/25
Trading Strategy Performance 0/25

My approach to calculate annuazlied returns, annuazlied volatility and sharpe ratio appears to be correct. I make a rough implementation just to check if I could clear the grader. Unfortunately, it did not. It could be an issue with the grader, I will wait a while before attempting this part again.

Data visualization with t-SNE (80/100)

Current submission
Moving Average 20/20
Linear Regression of Stock Returns 20/20
Unexplained Log-Returns 20/20
Project PCA 20/20
Visualize T-SNE 20/20

It appears to me that Project PCA is having issue with the grader.
This particular part of the assignment appears to be a very simple KernelPCA fit transform of df_test. At the same time, the index replication plots appear to be reasonable with PCA_1 tracing the trendline of SPX reasonably. Hopefully, the staff will fixed this issue and I will update this assignment later on.

ML

Bank defaults prediction using FDIC dataset (80/100)

Current submission

Logistic Regression Stats Model 20/20
Logistic Regression using Scikit-learn 20/20
Logistic Regression fewer predictors 20/20
Logistic Regression with Tensorflow 0/20
Logistic Regression with Neural Network 20/20

I am left with Logistic Regression with Tensorflow which did not clear the grader (your answer to Logistic Regression with TensorFlow is not correct). In case, there is any issue with the grader, I will wait for a while before attempting this question again. If anyone managed to clear this and would like to help, feel free to start a discussion.