[MUSIC] Hello, class. How's everyone today? Okay, first, let us review some jargon. Many student who have children to take this class have already studied basic statistics, mathematics, and financial economics. Maybe after taking our first course, there will be no end to the story of financial economics related to investment, but we start our lecture with one or two critical comment. There is only one reason to invest, right? It is to get a positive return. And if this rate of return is above bank deposit, it is only because it takes risk. No free lunch theorem applies here too. This is the most fundamental principle, which can be called the most basic modern investment theory framework. And the most fundamental theories of financial economics based on these principles are the modern portfolio theory and the capital asset pricing model. I'm sure a lot of you know, but I'll briefly mention it. Modern portfolio theory divides the risk of investment we have mentioned into systematic risk and idiosyncratic risk. Systematic risk is a common risk faced by many other assets and idiosyncratic risk creates a unique risk that only this asset has. And now, according to modern portfolio theory, idiosyncratic risk can be eliminated almost all through portfolio diversification. One more thing to mention is the capital asset pricing model, CAPM. CAPM demonstrate in one new formula the one factor, the market, or assets' return. Before we introduce other risk factors, we got this risk of systematic risk. And the degree of exposure to this risk is called beta. CAPM explains the market factor, but many scholars, investors later found that there were many other factors beside this. A good example is the Fama French three factor model. And each of these factors had factor beta, just like the market factor. A diversified portfolio can eliminate idiosyncratic risk. However, systematic risk by the factor, which is a common impact on most of asset, cannot be eliminated by creating of portfolio. That said, the beta of specific factor can be adjusted by changing the asset's allocation. Now, let us summarize the premise that we are putting on the basics. First, returns are reward for taking risk. Second, there is a risk that comes from a commonly applicable factor, such as systematic risk. For each asset, beta is different, since the impact from the systematic risk is different. Third, studying the factor that brings a risk by first and second can also link returns. We can come up with at least three conclusions like this. One thing, though, we are going to focus in this course is what factors are going to affect the investment returns. Of course, we go through the fund process to judge which factor is riskier, eliminate the idiosyncratic risk by combining assets, or making the overall competition of the systematic risk from different factors. However, in this lecture, we'll focus on finding out what affect the rate of return first. I'm going to introduce you to the method of modeling to figure this out, starting with the regression method that you know very well. And I'm going to introduce a variety of machine learning methods through connective lectures. I'll often use the word factors in my lecture Are you did to describe anything that can be related to rerun of assets. Since the early 1900, many investors have used the data to validate and provide their ideas in their own. Funds that invest using data and models are often called quant fund. And already in 1980s, many companies, now known as global quant fund, have started. The history of making investment decision using data and models is quite long. However, even in this long history, there has been an enormous trend in the last 10, 15 years. By analyzing data, creating model, creating algorithm to make a decision and automatically trading fast, it has taken over the financial market and at a rapid pace. Manufacturers have contributed to this change. But most of all, we cannot leave out the fact that computing power had become tremendous and affordable and data has become abundant. And with that, the models we use to take advantage of this computing power and data are more diverse than ever before. These models are what we'll talk about in this lecture. It didn't come out over anywhere, it was created through a very long time of research, and the quant fund mentioned earlier have long been conducting numerous studies. Computing power and technology have also dramatically changed the way we trade. Most of you familiar with the stock market may think the trading or financial asset is now at each using computers or cellphones. But it's not that long since financial asset beside stocks were traded in that way. The exchanges began to take place rapidly in the 2000s, especially since 2008. There have been many place called Electronic Communication Network, ECN, the place are similar stock exchanges where was in connect buyers and seller in the stock market. And this kind of transaction has become common in the bond and foreign exchange. With so much technology tied to the transition of investment and asset trading, institutional investors began to do a lot of direct market access, DMA. Direct market access, DMA, means having an infrastructure that connect directly to the exchange. By connecting to it right away, you can get to the exchange faster than other investors. This effort to have become what we call high frequency trading, HFT. HFT the form of a swift transaction per millisecond, many times in a very short time. Of course, humans cannot make a decision for such transactions. They are all models and algorithms. The investment strategies used by HFTs vary. Some of them were in the spotlight of financial regulators. This trend of investing using computer data and models have recently broken into the public at a very rapid pace. The past are affecting investment as much better. Earlier I said that the factor influence the return and allow us to predict the future returns. Several factors of high assets return, models and data analysis enable us to separate the factors and investors separately. It used to be only for Wall Street's big fund and institutional investors. Now these are all available to individual investors. The most important thing is that quite a few of fund established in the 1980s and invested consistently using data analysis and model performed better than the fund that made a decision based on investor's intuition over very long time. As you may have heard of those companies are Renaissance Technologies, DE Shaw, Citadel, Two Sigma, AQR, etc. There are a lot more successful players in the global financial market. The word of investment is comparative, maybe it is as much as beyond your imagination. Often, if there are people who make a profit, there are also people who make loss. And now to win this competition, data scientists in the investment sector make a tremendous effort. This tremendous effort can be seen in two main area. The first stage to obtain and process various data, they can be a source of various information. The second is to use many different models. These are two the main ingredient of our lecture. First, we all agree that the reason investors do this is to make a profit. Of course, we try hard to avoid risk, but let us focus on one direction, return. When we talk about returns, we always talk about two categories of return. I already mentioned those, beta and alpha. Probably many student know what these two words are, but since it is lecture, I'll explain them. Beta is precisely how much exposure your investment has to the factor. So it is often used to talk about the return you get from exporting the factor that bring you both return and risk. Alpha means the return that does not come from the exposure of the factor. The people who use data and model to invest can tell exactly how much return come from beta and how much alpha they are. This is possible because we use models. We'll discuss this farther in the lecture later. Funds that have made a huge return over the years mentioned above to everything to find new alpha sources. Again, we find the new information, analyze it and create algorithm for trading using modeling. In this process, alternative data machine learning method, which have recently received a lot of attention, are used. Speaking of which, it seems like this had happened lately, but it has been going on longer than you might think. Of course, some alternative data has only been available in the last decade or so. So before we talk more about machine learning and data in this class, let's briefly define what machine learning is and what alternative data is. Machine learning is an AI application, which means a computer program that use data and learns from data. Machine learning started from collecting data from what has already happened, finding pattern from the data, and making decision based on what has been learned from data and pattern when something happens in the future. It doesn't sound very easy, but it is simpler than you think. Using historical data, we figure out how investment returns are determined, and the rule we figured out will be expressed as a mathematical formula. Now, in the future, when the new information we used to define the pattern of return is generated, it goes into the formula we created and we make a decision based on that outcome. But one thing here is, this process must be updated continuously. When a new future occurs, the algorithm learns and updated, and this new algorithm is again used to make future decisions and it is to continue the process of learning again. We'll talk about how investors use data and model to do this. Then what is alternative data? Alternative data include satellite image, credit card sales, mobile geolocation data, and website scrapping, etc. You can use this data to make a decision through [INAUDIBLE] much anonymous set. The outcome from the model to make investment decisions, buying and selling asset, we call trading signal. These are generated from machine learning models' formulas. Today, I talked about what I will study with you from now on. We'll talk about many models through lectures for many hours from now on. [MUSIC]
