1. Statistics of the self-organized clusters shwoing the model variables of the screened stocks
3. Short/Long Areas of the self-organizing map.
Though generated for serious analysis of the U.S. stock markets, these colorful self-organizing maps are attractive in an artistic way. Taken from my financial markets Blog, these images were generated using Self-Organizing Map technology [SOM], which is a form of Artificial Intelligence. In essence the map clustered elements according to their degree of similarity. Imagine it as a matrix with each row representing a stock, and each column filled with data on fundmental and technical characteristics of the stock e.g. its P/E, the last 5 years Return, the Sharpe Ratio, the estimated Earnings Per Share etc. All in there were about 30 variables for each stock. The SOM took into account all the variables of all the stocks simultaneously when plotting the Map- a feat that no human brain can achieve or visualize. Think of it as a hugely multi-dimensional problem- if one row and three columns represents three dimensions, then in this case 600 rows and 30 columns represents a dimension of m rows x n columns
The output from three screens for Value stocks, Growth stocks and Quality stocks was plotted on the map. Since there was a Long and a Short version of these screens altogether there were about 120 stocks. Add to this, the 500 stocks of the S&P500 Index. Screened stocks are labeled with their Type (G=Growth, V=Value, Q=Quality), mode (L=Long, S=Short) and Sector to which they belonged [B=Basic Industries, C= Capital Goods, D=Consumer Durables, E=Energy, F=Finance, H= Health Care, ND=Consumer Non-Durables, T=Technology, TP= Transportation, U=Public Utilities]. Thus GLE is a Long Growth stock from the Energy sector or VSH is a Short Value stock from the Health Care sector. I= component of S&P500 Index.
Image 1 shows three clusters. Yellow clusters contain most of the stocks with an L i.e. long. Red cluster contains most of the stocks with an S i.e. Short. Blue cluster contains most of the S&P500 Index components. Note that 'I' stocks are also present in Yellow and Red clusters and a sprinkling of screened stocks are scattered over the Blue cluster too.
Image 2 with all the colorful bars shows some of the fundamental attributes of the the three clusters, S1, S2 and S3. The length of each bar is a measure of the deviation of the cluster mean from the Mean of the entire data set. The coded color key of each attribute is at the bottom of the chart.
Image 3 shows that after interpolation by the SOM based on the screened stocks, the map is divided into Long area [Red], and Short area [Blue]. Ambiguous areas are denoted as non-Red and non-Blue i.e. Green, Yellow, Organge]