This project’s objective is to define directionality of the closing price of a stock/ETF N days in the future relative to its daily opening price i.e. whether its price is predicted to increase (up), or decrease (down). Given the binary nature of the outcome, supervised learning techniques will be used to achieve the objective; specifically, classification techniques such as Decision Tress, Gaussian Naïve Bayes, Support Vector Machine and Neural Networks (in order of complexity) will be used. It used sensitivity analysis to define best parameter set for maximizing prediccion accuracy (AUC_ROC)

This project use the following datasets for ETF and Indexes:

• “Daily and Intraday Stock Price Data” from Kaggle.com(https://www.kaggle.com/borismarjanovic/daily-and-intraday-stock-price-data). It contains full historical daily (an intraday) price and volume data for all U.S.-based stocks and ETFs trading on the NYSE, NASDAQ, and NYSE MKT (Date, Open, High, Low, Close, Volume, OpenInt). This data set is adjusted for splits and dividends. Specifically, the daily data will be used.

• NASDAQ-100 Technology Index (NDXT) https://finance.yahoo.com/quote/%5ENDXT/history?p=^NDXT

• Volatility Index (VIX) http://www.cboe.com/products/vix-index-volatility/vix-options-and-futures

• NASDAQ Volatility Index (VXN) http://www.cboe.com/products/vix-index-volatility/volatility-on-stock-indexes/cboe-nasdaq-100-volatility-index-vxn

• List of best performing ETFs (http://etfdb.com/etfdb-category/technology-equities/%23etfs&sort_name=ytd_percent_return&sort_order=desc&page=1)

Engineered Features:

• Up_Down: defines whether the stock increased or decreased its price each day
• ETF’s Price Momentum
• ETF’s Price Volatility
• Sector’s Momentum - NDXT
• Sector’s Volatility - NDXT
• Market’s Momentum - VIX
• Market’s Volatility - VIX
• Market’s Momentum - VXN
• Market’s Volatility - VXN

These datasets will be combined into one dataframe for easy training and testing afterwards. This is, data from VIX, NXDT and VXN data sets will be brought in into one dataframe. The dataset contains 38921 rows for 13 “Tickers” with a data range from 2/23/2016 through 11/10/17.

It uses the following libraries which required no special installation procedure:

import numpy as np import pandas as pd import random import datetime as dt from pandas.core import datetools import time from matplotlib import pyplot as pyplot import csv import os

from sklearn.preprocessing import StandardScaler from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import FunctionTransformer

from sklearn.model_selection import GridSearchCV from sklearn.model_selection import TimeSeriesSplit

from sklearn.metrics import accuracy_score from sklearn.metrics import roc_curve, auc,roc_auc_score

from sklearn.neural_network import MLPClassifier from sklearn.svm import SVC from sklearn.gaussian_process import GaussianProcessClassifier from sklearn.tree import DecisionTreeClassifier from sklearn.ensemble import RandomForestClassifier from sklearn.naive_bayes import GaussianNB

from plotnine import *

• MLND_Capstone_Stock_LP_2.ipynb: contains code and exploratory data analysis, results and comments. Written in Python (2.7.12) in Jupyter Notebooks
• 20180310 Capstone Report.pdf: project report, containing analyses and following project rubric items
• 20180201 Proposal.pdf: Approved proposal for Capstone Project