Stock-Prices

Two Datasets were used:

historical_stock_prices
historical_stocks

Objectives

This study intends to make a forecast prediction of the adjusted closed price of several companies. First of all, we intend to analyze the dataset, evaluate the sectors of the predominant companies on the market as other factors as volume stocks, open prices.. all features that can be possible correlated with what we want to predict. In adition, if necessary we intend to enrich the dataset with all the information we can find that will help us to improve stock forecasting for certain companies.

Subsequently, use several traditional machine learning models to predict these adjusted closed prices such as linear models, decision trees, SVM... Secondly, use Deep Learning algorithms, namely RNN and LSTM/GRU to compare results with the first approach.

Finally, tune our models and draw relevant conclusions.

historical_stocks

	ticker	exchange	name	sector	industry
0	PIH	NASDAQ	1347 PROPERTY INSURANCE HOLDINGS, INC.	FINANCE	PROPERTY-CASUALTY INSURERS
1	PIHPP	NASDAQ	1347 PROPERTY INSURANCE HOLDINGS, INC.	FINANCE	PROPERTY-CASUALTY INSURERS
2	TURN	NASDAQ	180 DEGREE CAPITAL CORP.	FINANCE	FINANCE/INVESTORS SERVICES
3	FLWS	NASDAQ	1-800 FLOWERS.COM, INC.	CONSUMER SERVICES	OTHER SPECIALTY STORES
4	FCCY	NASDAQ	1ST CONSTITUTION BANCORP (NJ)	FINANCE	SAVINGS INSTITUTIONS

We realize there are 5 columns and this dataset:

ticker corresponds to the name of the share
exchange corresponds to the type of exchange made
name refers the company's name
sector refers to the actual sector where the given company operates
industry specifies the type of services that can be provided

historical_stock_prices

	ticker	open	close	adj_close	low	high	volume	date
0	AHH	11.50	11.58	8.493155	11.25	11.68	4633900	2013-05-08
1	AHH	11.50	11.55	8.471151	11.50	11.66	275800	2013-05-09
2	AHH	11.50	11.60	8.507822	11.50	11.65	277100	2013-05-10
3	AHH	11.50	11.65	8.544494	11.55	11.60	147400	2013-05-13
4	AHH	11.50	11.53	8.456484	11.50	11.60	184100	2013-05-14
...	...	...	...	...	...	...	...	...
20973884	NZF	14.60	14.59	14.590000	14.58	14.62	137500	2018-08-20
20973885	NZF	14.60	14.58	14.580000	14.57	14.61	151200	2018-08-21
20973886	NZF	14.58	14.59	14.590000	14.57	14.63	185400	2018-08-22
20973887	NZF	14.60	14.57	14.570000	14.57	14.64	135600	2018-08-23
20973888	NZF	14.60	14.69	14.690000	14.59	14.69	180900	2018-08-24

This dataset contains 8 columns:

ticker corresponds to the name of the share
open describe the open price of that share in a specific day
close describe the final share price in the end of a day
adj-close it´s a tricky column, describes the ajudsted price of a share, thats normally different from the close price
- An example of this, is when a stock splits occur. A stock split it's a current way used for compannies to sell more stocks, by diving the price in (x), lets say x = 2, then if one share = 10€, then, when stock split occurs, let say with a split=2, the share is equal to 5€, but in the end this 2 shares represent the same as 1 share, e.g, imagine that the companny have 10 shares, if you buy 1 share you have 1% of the company, in a stock split(split=2), if you buy 2 shares you only have 1% of the comapnny two.
- A stock's price is typically affected by supply and demand of market participants. However, some corporate actions, such as stock splits, dividends / distributions and rights offerings can affect a stock's price and adjustments are needed to arrive at a technically accurate reflection of the true value of that stock.
low is the lowest value paid for that share
high is the highest value paid for that share
volume represents the shares purchased in that day
date represents the date (year-month-day)

Machine Learning Algorithms

Linear Regression
Support Vectors Regression SVR
Decisions Tree / Random Forest
Recurrent Neural Networks (RNN) / LSTM

Group

André Guilherme Nunes Viveiros - a80524
Guilherme Marques Andrade – a80426
Rodrigo Teixeira Pereira – pg38424

pereira18 / Stock-Prices