This is a final class project for INDENG 185: Emerging Tech and Social Impact Challenge Lab at UC Berkeley. More information can be found on our slide deck: https://docs.google.com/presentation/d/1X0IET0fqcf-If5DrcH0cXRWoBmhCp4HgbDPMjGMy1H4/edit#slide=id.ga11bca5b12_0_0

This repo contains the Portfolio Optimization feature of Retiredge, implemented and annoated through Python and Jupyter Notebooks. It also contains a sample ETF data retrieved from Quandl, which will be used as sample for our predictive. Lastly an image is included as the Mobile GUI prototype to give insight on our final application. The process utlizes Efficient Frontier Calculation to suggest optimized portfolio budget allocations for users. The process is efficient and requires minimal financial knowledge to understand, perfect for our target market of college students.

The goal of this implementation is to show how data analysis and machine learning algorithms can be feasible performed on financial datasets to help our users make informed decisions. There are many spaces open for modification and additional intergrations, some potential additions include:

• Personalization
  • Through Machine Learning on the User history and Financial information
    • The investment style of a user (aggressive or conservative)
    • The user perference for certain funds/stocks
  • The degree of freedom in final portfolio decisions (how much the user relies on the ML predictions)
• Diversivied Prediction
  • Including more diverse data for prediction and analysis
    • This includes data beyond market data: e.g. Social data through News and other Media
  • Unsupervized learning can be performed on these datasets to identify more hidden trends

In terms of integrating with the actual application, only a small fraction of the displayed information will be included in the GUI through the Portfolio Recommendation interface. Much of the functionalities and processes will be hidden to the user for a more streamlined and friendly experience. However, even with a simple user interface, an incredible amount of backend analysis can be performed for an optimzed investing experience.

The code sample can be viewed directly through github without any installation. However, to modify and edit the file, jupyter notebook is needed.

pip install jupyterlab

The methods and functions can be used on a variety of financial datasets, which can be imported locally with the following code inside the notebook.

df = pd.read_csv('ETFG_FUND.csv')

Alternatively, data can be retrieved from financial platforms such as Quandl. Which can be installed and used with the code:

pip install quandl

An API key can be retrieved for free through creating an account on Quandl.

import quandl

quandl.ApiConfig.api_key = '##################'
stocks = ['ETFG/ANLT']
data = quandl.get_table('WIKI/PRICES', ticker = stocks,
                        qopts = { 'columns': ['date', 'ticker', 'adj_close'] },
                        date = { 'gte': '2016-1-1', 'lte': '2017-12-31' }, paginate=True)

Pull request are not welcome as this is a final class project.