coding and training bot predicting gold prices using machine learning (SVM) and Quandl

Question

I would like to build a bot that could predict gold prices using machine learning.

I started looking for factors that could potentially affect gold prices such as oil prices, stock etc. for the past 10 years for training data.

I thought Quandl would be the best bet for financial data of previous years.

How do I start coding and training the bot?

Answer 1

Your first step would be as you say to gather the data you need to train your model. I haven't used Quandl, but you can either download it and save as a.csv or you can query it through an API if Quandl has that.

Once you have some data to train your model on you'd want to look into LSTM.networks. If you have never worked with machine learning before start with a simpler project as LSTM's can be a bit daunting for complete beginners.

You wrote SVM in your headline, I assume you're referring to support vector machine. SVM is most commonly used for classification and most of the papers, examples and tutorials I have read that are similar to your problem use systems like LSTM as they perform better at these types of problems.

Answer 2

The first step would be gathering data and selecting the features which actually affect the gold price. Try to avoid any irrelevant feature, as it will going to effect the accuracy of the model which you will going to build.

I guess you are a newbie in Machine Learning, so i would suggest that instead of trying to get perfect results, start implementing some basic algorithms like regression. Although they would not give high accuracy, but they can provide you a good start.

It is always preferred to have a good knowledge of the algorithm that you are using. But you can use direct implementation in sklearn.

A link for reference http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LinearRegression.html