News history for stock prediction
In a previous post, I described a method (called “Newstrace”) to analyze the impact of news articles on stock prices by comparing the daily changes on days with similar news with other days. The main idea is that similar news may have the same impact on the price of a stock. Looking at a new news article you could use the information on the impact of similar news in the past (“Newstrace”) as a signal for trading decisions.
The steps of this method could be done with some software tools and python scripts as described in my article. But it takes a lot of work to analyze a single text to make the right conclusions. And there are thousands of relevant news each day. So I developed a new service called “Pinalyze” (www.pinalyze.com) which could do the work for me.
Machine learning is a branch of artificial intelligence that has grabbed a lot of headlines previously. People are…www.datadriveninvestor.com
Pinalyze reads the news from important publishers and makes the steps of the analysis immediately after the news was published for each article. At the moment of the writing 17 different news sources like Bloomberg, CNN, etc. are covered. The analysis is done for the stock prices of several companies (209 at the moment).
Pinalyze then selects from the huge amount of information it collects the relevant information for you. Defining the stocks and the news sources you are interested in, you get a list of news with trading signals which may be interesting for you.
After registering for the free plan you can define the profile of your interests and look at the list of signals the system found in the last days.
Articles with positive sentiment are in green, negative ones in red with small arrows pointing up or down indicating the direction of movement of the stocks.
If you don't want to miss a new signal you can check the “ Send me mail reminders” checkbox in the profile settings to get an email with the basic information of the signal and a link to the analytics dashboard. From the website, you can look at the analysis dashboard by clicking at the headline or the “More Infos” Button. For this page, you have to subscribe to the Premium plan.
The dashboard shows the first sentences of the article and gives an overview of the stocks with interesting movements up or down on similar past articles. You can inspect the list of these related articles by clicking a button getting a list with links to the original news pages and a green or red dot indicating the sentiment of the past article.
The button “Analysis” brings up an analysis of the statistical properties of the signal and a plot of the distribution of the daily changes for days on the newstrace versus the rest of the days in the same timespan.
The graph compares the distributions for daily changes on the “Newstrace” (days with news similar to the analyzed article) with the rest of the days.
- Is the “Newstrace” distribution and it's mean shifted to the left or right?
- Are the distributions normal distributions?
- Are there any outliers in the data?
You can also look at the line chart of the stock prices in the inspected timespan with green and red bars indicating the publishing days of the “Newstrace” articles. The price movement on these days is marked with green/red flags and the sentiment of the articles with dots.
A chart of the histogram of the changes and a boxplot can help you going deeper into the statistical analysis for example to find trends, outliers in the data or to visualize the shape of the distribution.
In my next post, I will show some of these steps on an example news article having some signal.