Having previously dealt with sentiment analysis of user reviews and studied a bit of quantitative investment, I recently came up with an investment strategy. Although the practice has not been carried out, but from the point of view of logical derivation, it is very appropriate.
I want to share this investment strategy with you through this article, to provide you with a new perspective on stocks. At the same time, we also hope to get everyone’s opinions and feedback to make up for the blind spot.
China’s stock market is a market dominated by retail investors, which is easily affected by news and sentiment, and will directly reflect the rise and fall of stock prices. A good news is likely to quickly lift a stock, while a bad news is likely to lead to a rapid decline, that is, chasing the rise and killing the fall.
And how does the source of this kind of good news spread?
- First, the public announcements and financial statements disclosed by listed companies are the truest and most stable source of information.
- Second, the analysis of various investment institutions, while very effective, is extremely difficult for retail investors to obtain.
- Finally, numerous investment field opinion leaders (BIG V), published articles, short messages, etc.
It is very difficult for individual investors to read the announcements and financial results of listed companies, requiring strong financial expertise and a lot of time and energy. As a result, very few retail investors read company statements for themselves to make investment decisions.
For the average retail investor, institutional analysis is neither readily available nor intelligible. Thus, this information pathway is also blocked.
To sum up, the most common way for retail investors to get news is to judge and invest in stocks by following big Vs and reading their articles and short messages. At the same time, the reason why big V is called big V must have its professionalism, or familiar with the industry, or understand financial statements.
So, it is conceivable that the investment advice of the big V will certainly affect the investment decisions of the majority of retail investors, and there will be a short and medium term, more certain income expectations.
However, there are two problems with such earnings expectations for the average retail investor:
- The delay of information, the transmission of information through social networks, there is a delay. The same message, not all retail investors can read at the same time, there must be a time difference.
- Information conflict, the same news, different big V may have completely different interpretation, which will lead to conflicts, let retail investors make different choices.
These two problems will produce a certain degree of information difference, also gave birth to speculative opportunities. If we can develop a system that prioritizes the data of big Vs ahead of the vast majority of retail investors and gets in early, then we can enjoy the dividends of poor information.
As for the specific steps of this quantitative strategy, I have roughly sorted them out:
1. Collect big VS in the investment field through platforms such as Snowball, Weibo, wechat and Zhihu
In the comprehensive community, there will be many big Vs talking about stocks. We can find them through Weibo, wechat public account, Zhihu and other platforms, and record their user information (user name, user ID).
Snowball, as a vertical community in the field of stock investment, is worthy of our attention.
We can develop a standard to assign different weights to different platforms and big Vs, so as to more accurately measure their influence, such as the number of fans, the number of interactions, the frequency of posts, etc.
Thus, we have a weighted list of big V’s as targets for our climb.
2. Regularly climb V articles, short messages and other information
If the corresponding platform provides relevant API interface, even if it is paid, it is better to use API interface, mainly because its stability is guaranteed to avoid crawler being shielded or invalid. After all, this involves quantitative investment and requires high timeliness and stability.
In this way, we can complete large-scale and efficient data collection, and get first-hand information of each big V in the first time, far ahead of the vast majority of retail investors.
3. Analyze the ups and downs of stocks in the text
Although the analysis here is the rise and fall of the stock, but the principle and text sentiment analysis is the same. I wrote an article about emotion analysis before, “Talking about how to do emotion analysis”, which explains the method of emotion analysis, we can take a look.
We need two sets of dictionary data, respectively:
- Dictionary of stock names and codes
- Words related to bulls and bears
Stock name and code
Through the stock name or stock code, it helps us to screen out the text fragment discussing the stock and screen out some irrelevant information, such as big V’s daily and life perception.
It also allows quantitative trading systems to focus on specific stocks and automate subsequent trades.
Special attention should be paid to the unification of stock names, because many stocks have different names, such as Moutai, Kweichow Moutai, etc. This maximizes matching of the right stocks and avoids omissions.
Bullish bearish words
This is similar to sentiment words in sentiment analysis, which are used to express attitudes towards something. The bullish and bearish of stocks will have some unique words, which need to be sorted out manually. Here are some simple examples:
Bullish vocabulary:
- good
- dividend
- Value plays
Bearish words:
- A bearish
- short
- overvalued
We can also assign a weight to these words to measure how much “emotion” they reflect, making it a better predictor of ups and downs.
With these two dictionaries, word segmentation is involved in the text, and the quality of word segmentation directly affects the final analysis effect, is a very important link. I have written a gold digging book “In-depth Understanding of NLP Chinese word segmentation: from principle to practice”, which explains the NLP Chinese word segmentation technology in detail, you can have a look.
At this point, we can extract the stocks involved in the text and synthesize the views of all the big Vs to generate a bullish stock list. At the same time, you can set a threshold to keep only those stocks that are better bet and that are favored by more big Vs.
4. Back test and simulation disk test
Although we already have a bullish list of stocks, don’t rush to invest, after all, it is real money, or very cautious.
Now, there are many quant trading platforms that offer backtesting. Therefore, we can use the historical data of BIG V and carry out back test in combination with the historical transaction data to see whether our strategy is effective and effectively discover the problems existing in the analysis.
When retesting, we also need some other investment strategies, to do a good job in capital management, but also to set a good stop loss point and stop profit point, because it is impossible to guarantee that every time will rise, but the probability of bullish is much greater than the probability of bearish, so, need to bear a certain risk of retreat.
If all goes well, then you can test your strategy on a mock board, where you trade in real-time at the same time, to ensure that your strategy is timely.
Backtesting is like the simulation environment of Internet companies, which is tested by internal testers, and the simulation disk test is a small traffic test.
5. Online quantitative trading strategy
As people will be affected by psychological effects, such as: fear, greed, loss aversion, fluke psychology and so on. So, once the strategy has been proven to work, you need to be bold enough to leave trading operations to machines.
The most important thing is to do a good job in capital management, and after the expiration of the strategy, the evaluation and optimization of the strategy.
This is a very important point, because a lot of people fail in stocks because they are too emotional, and this strategy takes advantage of that.
conclusion
Through the imagination of this quantitative strategy, we can find that NLP emotion analysis can do a lot of things, such as: when pursuing goddess or male god, we can analyze his/her Microblog and wechat, and timely analyze when he/she is prone to depression and needs comfort.
Most of the time, technology empowerment will form a dimension reduction blow, for ordinary people, there is almost no chance to resist. Therefore, if we want to become the rule makers of the game, we need to constantly improve ourselves, complete the transformation.
Finally, we have a book entitled “Understanding NLP Chinese Word Segmentation: From Principle to Practice”, which will help you master Chinese word segmentation from scratch and step into the door of NLP.
If because of the above content for your help, please help to point a like, comment, transfer hair, thank you very much!