It is often just a misnomer – the observed Hendrik Leber, co-founder of the German Fund house Acatis, if funds advertise with the term “artificial intelligence” (AI). To “Finance and economy” he declared: “There’s a lot of it is grown in a pot, for example, Big Data or Robo Advice.” Big Data is the processing of huge amounts of data. Under Robo Advice is the creation of automatic Portfolio-proposals for investors. “Besides, ordinary statistics is behind it often,” he adds liver critical.
the application of AI technology (see text box) to a complex System like the financial markets are promising. With the pattern recognition based on machine learning can be used to process large amounts of data and investment strategies to optimize.
a long time computer scientists and financial experts are working to make computers better investment decisions: In the eighties the first studies on the prediction of stock prices by KI-were published methods. However, as in other areas – such as voice and image recognition – the technology in the financial sector in recent years due to advances in software and Hardware travel.
“today’s computing power enables the development steps that were in the year 2000, only under the greatest effort possible,” observed Michael Günther. He’s cemented a portfolio Manager and a developer in the plant house. His Fund Trycon is based on trend following strategies and used since 2013 AI-methods. Günther says, for his Fund of the effort in data and computing power beyond a conventional trend-following strategy to the one hundred thousand times. The AI strategy could not be found to recognize Connections between data, “the traditional models.”
Christian Lopez, Research Director of Amundi, a subsidiary of CPR Asset Management, stresses the help of AI: “It is impossible for a human to look up all the news that happen in a Moment. Here Artificial intelligence can support – for example, the natural language processing – the Manager.” Also AI could be the human expert with investment proposals, assist.
The decision must still make the man: a Trader on the New York stock exchange. Photo: Keystone
Annette, She works with your company Xanadu alpha, a new investment approach to AI-based, and stresses: “Artificial intelligence is important to find patterns in huge amounts of data. But the decision, what pattern do we really believe, has yet to meet the man.”
Günther relies on the “semi-automatic” implementation of the decisions of the computer. The man looks at them again. According to Günther, it is important that “risk management is conventionally – in that the limits of the machine cannot override.”
The Fund house Acatis has three products in the range which automatically implement AI decisions. Two of them are together with the expert Jürgen Schmidhuber, Professor in Lugano. Acatis-portfolio Manager Hendrik Leber explained: “in addition, we use KI in a Fund to make a preselection of shares. The cooperation of man and machine, works well.” The Fund, with the Autonomous AI product but ran really well this year – “although you are not based at all on Momentum rather on fundamental factors,” says liver. “But we will still need some years of experience to the Performance-to assess the benefits of the approach.”
Hedge Funds are using AI solutions. According to a statistics of the data provider Eurekahedge, such approaches have brought about significant Performance advantage (see graphic). Stefan wall rich wall rich Wolf Asset Management is satisfied with his KI-model: “Last year, the baptism of fire was: in Spite of the decline in the stock market, we achieved a Performance of around zero.” He is active in standstill strategies – options are sold and premiums are taken.
wall rich explains: “Our approach is born out of the idea, the Performance of the different strategies to back-calculate. Our programmer had the idea that we do not expect, but that we optimize the strategies.” The machine could a chosen strategy be better than a human Trader – “you emotion loser”. Constantly be algorithm search of a better model that will be adapted every two to three years.
In the case of wall rich are price and volatility data of the Input. Michael Günther for the day is taken from a variety of market data as an investment decision. Leber of Acatis explains: “We want to create long-term forecasts on the basis of fundamental signals.” Thus, neural networks were (see text box below) can be used to find out the “unstructured data” such as company reports, relevant statements, which were not taken into account in the course.
“there is a completely new event in the markets, then our model is not able to provide that.”Hendrik Leber, Acatis-portfolio Manager
liver believes that in ten years, “KI is a dominant factor in Asset Management will be”. And more and more computing power will accelerate the pace of development. Michael Günther does not believe that trading strategies are not cannibalizing each other, when more providers move to the use of AI: “The models may differ in a variety of ways – for example, what data base is used.”
Annette, She stresses: “If the same data is looked at with the same glasses, is to get an excess Return to the market, Alpha, only on the edges.” Instead of “Data Mining” – that is, the mass analysis of huge data pools to find the preferred “Data Fracking”: a small, specific collections of data and evaluate.
Nobody knows cut, how well the AI models in the future. When someone say that his model did overcome a crisis in the past, wonderful, then it is likely to have been trained with data of that time, says the liver. He stressed: “there is a completely new event in the markets, and our model is not intended to provide for.”
keywords: artificial intelligence
the methods are grouped under the term artificial intelligence (in English: Artificial Intelligence) are numerous. In the foreground is the machine Learning: methods to detect, from the given input data as independently as possible, patterns, categorize, and draw conclusions.
hit the headlines in the past few years, particularly advances in artificial neural networks. They are used today for the detection of human language, Translations, proposals for the Internet Shopping, and the analysis of images.
The principle is the human brain – neurons are modelled on nerve cells, and has already been more than seventy years ago designed. The weights of Connections in a network of neurons will change, so you lead a combination of input data to the appropriate Output.
The neural networks “learn” from training data and change to find the optimal combination of Input – output data. Today’s diverse applications, but thanks to breakthroughs in recent years.
become, on the one Hand, the networks of “deep” – this means that several layers are used to neurons, with a very large number of these virtual nerve cells. Also, the training data are much more comprehensive: Google has used for a neural network for image recognition, hundreds of millions of images. This requires more computing power. In addition, new structures of neural networks are providing more and more impressive results of this “Deep Learning”.
The results of neural networks are often viewed as non-transparent: “This is a correct criticism”, explains Christian Lopez of CRP Asset Management. “The models can include distortion – that depends on the data used.”
Neural networks belong to the field of classifiers. These are meant to structure data in a way that makes sense. A further known method for decision trees. This classification can be in the Form of branches clearly. To make the most of uncertain and incomplete information conclusions, Bayes used networks. A further area of KI are evolutionary Algorithms. Here, computer models can be changed randomly, as is the case of biological Evolution, so that you can better solve the Problem.
Many data points are a blessing and a curse. Because the System should find such a perfect Portfolio, it can do this for the past great – but this is no guarantee. Hendrik Leber of Acatis explains: “If one has thousands of data, such as ten, which can be used to recreate a Course.” (Financial and economic)
Created: 14.03.2019, 12:59 PM
