Conference PaperPDF Available

Knowledge Discovery in Quarterly Financial Data of Stocks Based on the Prime Standard using a Hybrid of a Swarm with SOM

Abstract and Figures

Stocks of the German Prime standard have to publish financial reports every three months which were not used fully for fundamental analysis so far. Through web scrapping, an up-to-date high-dimensional dataset of 45 features of 269 companies was extracted, but finding meaningful cluster structures in a high-dimensional dataset with a low number of cases is still a challenge in data science. A hybrid of a swarm with a SOM called Databionic swarm (DBS) found meaningful structures in the financial reports. Using the Chord distance the DBS algorithm results in a topographic map of high-dimensional structures and a clustering. Knowledge from the clustering is acquired using CART. The cluster structures can be explained by simple rules that allow predicting which future stock courses will fall with a 70% probability.
Content may be subject to copyright.
Knowledge Discovery in Quarterly Financial
Data of Stocks Based on the Prime Standard
using a Hybrid of a Swarm with SOM
Michael C. Thrun1
1- University of Marburg, Mathematics and Computer Science
Hans-Meerwein Str., 35032 Marburg, Germany
thrun@deepbionics.de
Abstract. Stocks of the German Prime standard have to publish financial reports
every three months which were not used fully for fundamental analysis so far.
Through web scrapping, an up-to-date high-dimensional dataset of 45 features of
269 companies was extracted, but finding meaningful cluster structures in a high-
dimensional dataset with a low number of cases is still a challenge in data science.
A hybrid of a swarm with a SOM called Databionic swarm (DBS) found
meaningful structures in the financial reports. Using the Chord distance the DBS
algorithm results in a topographic map of high-dimensional structures and a
clustering. Knowledge from the clustering is acquired using CART. The cluster
structures can be explained by simple rules that allow predicting which future
stock courses will fall with a 70% probability.
1 Introduction
Some of the companies listed in the Frankfurt stock exchange comply with a
rigorously defined transparency standard higher than the general standard of stocks
traded on the world market. This market segment of stocks is called German Prime
standard and requires companies to publish their accounting information [1].
Currently, the Prime standard contains 324 companies [2]. Usually, fundamental
analysis has the purpose of predicting the market value of a stock. Without applying a
forecasting method, the “fundamental“ value of a stock is determined [3] and
compared to the current value. Some empirical studies have used accounting
information to predict the future performance of a firm [3]. Often fundamental
analysis is a part of a larger system with the goal to select the right stock (e.g.,
CANSLIM system [4]) using a low amount of variables and a larger number of stocks
[5-7]. Thus, “research does not fully exploit the wealth of information contained in
general purpose financial reports but is outside of the primary financial statements”
[8]. In this work, high-dimensional structures are investigated which are characterized
by 45 variables describing the quarterly financial statement, income sheet, and cash
flow defined by the German prime standard. The information is extracted directly
from the web [9] by applying a self-made web scrapping algorithm allowing the
extraction of data for 269 companies. The dataset was extracted for the first quarter of
2018, and the stock courses were extracted for the second quarter of 2019.
2 Methods
After preprocessing (i.e., handling of missing values, normalization, and
decorrelation) of data the Databionic swarm (DBS) algorithm can be applied [10] on
269 companies where 43 variables are used. The algorithm consists of three parts.
First, the high-dimensional data is projected into a two-dimensional space using a
swarm which utilizes game theory, self-organization, and emergence as well as swarm
intelligence [10]. Besides the number of clusters and a single Boolean parameter
describing the type of structure, the DBS does not require any parameters to be set.
Here, the Chord distance (c.f. [11] p.49) is chosen for cluster analysis because
distributions analysis of distances indicates a bimodality (Fig. 1). Statistical testing
with Hartigan’s dip test [12] agrees that the distribution is not unimodal
(p(N=36046,D= 0.027821)<2.2e-16). Bimodality in the distance distributions serves
as an indicator that there are smaller intra-cluster distances and larger inter-cluster
distances resulting in the assumption that a cluster structure exists. Note, that in
general, all financial reports are quite dissimilar because there are no small distances
below 0.5 in Fig.1.
In the second part, the visualization of a topographic map is generated using a
simplified emergent self-organizing map [13] in order to visualize high-dimensional
structures. The topographic map accounts for projection errors because two-
dimensional similarities in the scatter plot cannot coercively represent high-
dimensional distances [14] and common quality measures of dimensionality reduction
methods require prior assumptions about the underlying structures [15]. The
topographic map combines a 3D landscape with hypsometric tints. Hypsometric tints
are surface colors which depict ranges of elevation [16] intersected with contour lines.
“Blue colors indicate small distances (sea level), green and brown colors indicate
middle distances (small hilly country) and white colors indicate high distances (snow
and ice of high mountains). The valleys and basins indicate clusters and the
watersheds of hills and mountains indicate borderlines of c lusters” [17]. In sum, this
visualization is consistent with a 3D landscape for the human eye enabling 3D
printing of a representation of high-dimensional structures; therefore one has a haptic
grasp and sees data structures intuitively enabling layman to interpret them [17].
In [18] it was shown that the visualized elevation between two projected points is an
approximation of the input-space distance D(l,j) between the two high-dimensional
points (here two companies). Voronoi cells around each projected point define the
abstract U-matrix (AU-matrix) and generate a Delaunay graph . For every projected
point all direct connections are weighted using the input-space distances D(l,j)
because each border between two Voronoi cells defines a height. All possible
weighted Delaunay paths between all points are calculated. Then, the minimum of
all possible path distances between a pair of points in the output space O is
calculated as the shortest path using the algorithm of [19] resulting in a new
high-dimensional distance which defines the distance of each two companies
based on their financial accounting. In this case, the connected structure type of DBS
clustering is chosen where the similarity between two subsets of data points is defined
as the minimum distance between data points in these subsets and the clustering
process is agglomerative (c.f. [10]): Let be the distance between two clusters c1 I
and c2 I, and let D(l,j) be the distance between two data points in the input space I;
then the connected approach is defined with .
Through inspecting the topographic map the central problem of estimation of the
number of clusters is solved by counting the number of valleys and the structure type
can be set. A dendrogram can be shown additionally. The clustering is valid if
mountains do not partition clusters indicated by colored points of the same color [10].
3 Results
The high-dimensional structures of the financial reports of companies are visible in
the topographic map in Fig .3 where three valleys can be identified. The number of
valleys seen in the topographic map lead to the choice of three clusters. Additionally,
large changes in fusion levels of the ultrametric portion of the Chord distance indicate
the best cut (Fig.2, left, y-axis). After the clustering process, the heatmap in Fig. 2
indicates that more similar points are inside a cluster (yellow) and more dissimilar
points are outside a cluster (red). The three main clusters computed by the connected
approach of DBS are separated by mountain ranges (Fig. 3) and have an average
intra-cluster distance of c1=0.97, c2=1.01 and c3=0.91 being located in the first mode
of Fig. 1. The points in the topographic map symbolize the companies and are colored
by the clustering. Two outliers can be identified. The heatmap agrees with the
topographic map that the clusters consist of more similar points inside a cluster than
outside the cluster. Applying the CART algorithm [20] to the clustering yields a
simple set of rules (Fig. 5). If “net income from continuing ops NIFCO < -264 and
“operating income or loss OIL < -58.3 then class 2 is defined. NIFCO defines the
after-tax earnings that business has generated from its operational activities. NIFCO
“is considered to be a prime indicator of the financial health of a firm’s core
activities[21]. OIL is the difference between revenues and costs generated by
ordinary operations and before deducting interest, taxes et cetera [22]. The extracted
rules for class 2 lead to the hypothesis that stock prices of companies in the second
cluster are overvalued and will fall in the next quarter. Using stock prices of Q2/2018
the hypothesis is verified under the assumption that the data of all companies is
available on the same date at the beginning of the respective quarter of a year. In Fig.
4 the class-dependent MD-plot [28] of the rate of return of stocks is shown. The
courses were compared at the last trading day of the first quarter against the last
trading day of the second quarter in 2018 with relative differences [24]. The rate of
return in Class 2 is significantly lesser than in Class 1: Class 1 has a shift equal to
zero with a Wilcoxon rank sum test (shift not equal to zero p(N=196, V=9512)=0.86),
class 2 a shift not equal to zero with a Wilcoxon rank sum test p(N=54,W=6792)
<0.001. Stock courses of eight companies (3% of class 1 and 2 and outliers) could not
be extracted from the web.
Fig. 1: MirroredDensity plot (MD-plot [28]) of the R package ‘DataVisualizations’
on CRAN [23] shows a bimodal distribution of Chord distances with the first
mode around 1 and the second mode around 1.6.
Fig. 2: Dendrogram (left), and heatmap (right) of the distances sorted by the
clustering using the R package ‘DataVisualizations’ on CRAN [23]. In the
heatmap, the smaller distances are in yellow belonging to a clusters and the
larger distances in red in-between different clusters (c.f. Fig. 1). The branches
of the dendrogram are colored by the first three clusters.
Fig. 3: The topographic map can visualize 43 dimensional, distance-based structures.
It shows three valleys - one major cluster with companies represented by
magenta points (N=199), one smaller cluster with companies in yellow (N=57)
and black (N=10) and two outliers (green and red).
Fig. 4: MDplot [28] of the rate of return in % calculated with relative differences [24].
The red line marks a rate of return of zero. Class 2 has 30% of stocks with a
rate of return above zero and significantly differs from Class 1.
Fig. 5: CART shows three distinct rules where outliers are incorrectly classified: “net
income from continuing ops (NIFCO>=-264), “operating income or loss
(OIL<-58.3) and “net income applicable to common shares (NIACS>= 5.5).
4 Discussion
The topographic map visualizes a 43 dimensional, distance-based structures in a 3D
landscape in Fig. 3. It showed three distinctive valleys leading to the hypothesis that
the dataset has three clusters. The topographic map in Fig. 3 is noisy because the
sample of data is small but the feature space is large, and there are no small distances
(see Fig. 1). However, the distributions analysis (Fig. 1) and the heatmap (Fig. 2)
indicate high-dimensional and distance-based structures of the data which can be
visualized by the topographic map (Fig. 3). These structures are reproduced in the
cluster analysis and a dendrogram. The heatmap could also indicate that cluster three
is a sub cluster of cluster one, yet the clustering is valid because mountains do not
partition clusters and the intra-cluster distances are smaller than the inter-cluster
distances (Fig. 2). The primary cluster (N=199) does not yield interesting insights
about the data which is typical for cluster analysis [25]. However, the extracted rules
(Fig. 5) are interesting for the second cluster (N=57) where NIFCO has to be lower
than -264 and (OIL) lower than -58.3. By understanding these two variables through
the two rules, it can be concluded that the stock prices of these companies will fall in
the next quarter. This prediction can be verified by the significant higher probability
of a negative rate of return of stock prices of companies in class 2 in Q2/2018
compared to class 1. In sum, 7 out of 10 companies in class 2 lost value on the stock
market during the second quarter. In comparison, the success rate at stock picking by
a hybrid AI system was reported with on average 55.19 to 60.69% [26], and experts
had a success rate worse than chance [27]. Thus, the clustering allows a data-driven
stock picking with a high chance of success for a short position.
5 Conclusion
This work presents an example in which the DBS algorithm made it possible to apply
cluster analysis to high-dimensional data where only a low number of cases exist.
Further examples and a comparison to common clustering algorithms as well as to
dimension reduction techniques are presented in [10]. The clusters can be explained
by simple rules allowing to select stocks for the next quarter by simple thresholds in
the data (selling first and buying later). Besides the choice of the number clusters and
a Boolean parameter describing the type of structure, DBS is parameter-free and can
be downloaded as the R package “DatabionicSwarm” on CRAN.
Acknowledgments
Gratitude goes to Hamza Tayyab for programming the web scrapping algorithm which
extracted the quarterly data.
References
[1] Prime-Standard. Teilbereich des Amtlichen Marktes und des Geregelten Marktes der Deutschen Börse für
Unternehmen, die besonders hohe Transparenzstandards [18.09.2018]; from: http://de utsche-boerse.com/dbg-
de/ueber-uns/services/know-how/boersenlexiko n/boersenlexikon-article/Prime-Standard/2561178.
[2] Gelistete Unter nehmen in Prime Standard, G.S.u.S., http://www.deutsche-boerse-cash-market.com/dbcm-
de/instrumente-statistiken/statistiken/gelistete-unternehmen. 2018, Deutsche Börse: Frankfurt.
[3] Abad, C., S.A. Thore, and J. Laffarga, Fundamental analysis of stocks by two
stage DEA. Managerial and
Decision Economics, 2004. 25(5): p. 231-241.
[4] O'Neil, W.J., How to make money in stocks. Vol. 10. 1988: McGraw-Hill New York.
[5] Deboeck, G.J. and A. Ultsch, Picking stocks with emergent self-organizing value maps. Neural Network World,
2000. 10(1): p. 203-216.
[6] Ou, J.A. and S. H. Penman, Financial statement analysis and the prediction of stock returns. Journal of accounting
and economics, 1989. 11(4): p. 295-329.
[7] Mohanram, P.S., Separating winners from losers among lowbook-to-market stocks using financial statement
analysis. Review of accounting studies, 2005. 10(2-3): p. 133-170.
[8] Richardson, S., I. Tuna, and P. Wysocki, Accounting anoma lies and fundamental analysis: A review of recent
research advances. Journal of Accounting and Economics, 2010. 50(2-3): p. 410-454.
[9] Yahoo! Finance. Income statement, Balance Sheet and Cash Flow. 2018 [cited 2018 29.09.2018]; Available from:
https://finance.yahoo.com/quote/SAP/financials?p=SAP (Exemplary).
[10] Thrun, M.C. , Projection Based Clustering through Self-Organization and Swarm Intelligence. 2018, Heidelberg:
Springer.
[11] McCune, B., J.B. Grace, and D.L. Urban, Analysis of ecological communities, chapter 6. Vol. 28. 2002: MjM
software design Gleneden Beach.
[12] Hartigan, J.A. and P.M. Hartigan, The dip test of unimodality. The annals of Statistics, 1985. 13(1): p. 70-84.
[13] Ultsch, A. and M.C. Thrun, Credible Vis ualizations for Planar Projections, in 12 th International Workshop on
Self-Organizing Maps and Learning Vector Quantization (WSOM). 2017, IEEE: Nany, France. p. 1-5.
[14] Dasgupta, S. and A. Gupta, An elementary proof of a theorem of Johnson and Lindenstrauss. Random Structures &
Algorithms, 2003. 22(1): p. 60-65.
[15] Thrun, M.C. and A. Ultsch. Investigating Quality measurements of projections for the Evaluation of Distance and
Density-based Structures of High-Dimensional Data. in European Conference on Data Analysis (ECDA).
2018. Paderborn, Germany.
[16] Patterson, T. and N.V. Kelso, Hal Shelton revisited: Designing and producing natural-color maps with satellite
land cover data. Cartographic Perspectives, 2004(47): p. 28-55.
[17] Thrun, M.C., et al., Visualization and 3D Pr inting of Multivariate Data of Biomarkers, in International Conference
in Central Europe on Computer Graphics, Visualization and Computer Vision (WSCG), 2016: Plzen. p. 7-16.
[18] Lötsch, J. and A. Ultsch. Exploiting the Structures of the U-Matrix. in Advances in Self-Organizin g Maps and
Learning Vector Quantization. 2014. Mittweida, Germany: Springer International Publishing.
[19] Dijkstra, E.W., A note on two problems in connexion with graphs. Numerische mathematik, 1959. 1(1): p. 269-
271.
[20] Breiman, L., et al., Classification and regression trees. 1984: CRC press.
[21] Bragg, S. Net income from continuing operations. 2018 [c ited 2018 03.11.2018]; Available from:
https://www.accountingtools.com/articles/2017/5/12/net-income-from-continuing-operations.
[22] Silver, C., et al. Operating Income. 2014 [cited 2018 03.11.2018]; Available from:
https://www.investopedia.com/terms/o/operatinginco me.asp.
[23] Thrun, M.C. and A. Ultsch, Effects of the payout system of income taxes to municipalities in Germany, in 12th
Professor Aleksander Zelias International Conference on Modelling and Forecasting of Socio-Economic
Phenomena, 2018, Foundation of the Cracow Univers ity of Economics: Cracow, Poland. p. 533 -542.
[24] Ultsch, A., Is Log Ratio a Good Value for Measuring Return in Stock Investments?, in Advances in Data Analysis,
Data Handling and Business Intelligence. 2009, Springer. p. 505-511.
[25] Behnisch, M. and A. Ultsch, Knowledge Discovery in Spatial Planning Data: A Concept for Cluster
Understanding, in Computational Approaches for Urban Environments. 2015, Springer. p. 49-75.
[26] Tsaih, R., Y. Hsu, and C.C. Lai, Forecasting S&P 500 stock index futures with a hybrid AI system. Decision
Support Systems, 1998. 23(2): p. 161-174.
[27] Torngren, G. and H. Mont gomery, Worse than chance? Performance and confidence among professionals and
laypeople in the stock market. The Journal of Behaviora l Finance, 2004. 5(3): p. 148-153.
[28] Thrun, M.C. and A. Ultsch, Analyzing the Fine Structure of Distributions. Technica l Report being submitted, Dept.
of Mathematics and Computer Science, Philipps-U niversity of Marburg, 2019: Marburg: p.1-22.
... Table 1 summarizes the interesting basic properties from the perspective of data mining and the methods used to compare the performance of different methods. Extensive knowledge discovery for this dataset was performed in [17]. Therefore, we compare the visualizations with basic descriptive statistics and show which visualization tools do not visualize the shapes of the PDF accurately without changing the default parameters of the investigated visualization methods. ...
Article
Full-text available
One aim of data mining is the identification of interesting structures in data. For better analytical results, the basic properties of an empirical distribution, such as skewness and eventual clipping, i.e. hard limits in value ranges, need to be assessed. Of particular interest is the question of whether the data originate from one process or contain subsets related to different states of the data producing process. Data visualization tools should deliver a clear picture of the univariate probability density distribution (PDF) for each feature. Visualization tools for PDFs typically use kernel density estimates and include both the classical histogram, as well as the modern tools like ridgeline plots, bean plots and violin plots. If density estimation parameters remain in a default setting, conventional methods pose several problems when visualizing the PDF of uniform, multimodal, skewed distributions and distributions with clipped data, For that reason, a new visualization tool called the mirrored density plot (MD plot), which is specifically designed to discover interesting structures in continuous features, is proposed. The MD plot does not require adjusting any parameters of density estimation, which is what may make the use of this plot compelling particularly to non-experts. The visualization tools in question are evaluated against statistical tests with regard to typical challenges of explorative distribution analysis. The results of the evaluation are presented using bimodal Gaussian, skewed distributions and several features with already published PDFs. In an exploratory data analysis of 12 features describing quarterly financial statements, when statistical testing poses a great difficulty, only the MD plots can identify the structure of their PDFs. In sum, the MD plot outperforms the above mentioned methods.
Article
Full-text available
One aim of data mining is the identification of interesting structures in data. For better analytical results, the basic properties of an empirical distribution, such as skewness and eventual clipping, i.e. hard limits in value ranges, need to be assessed. Of particular interest is the question of whether the data originate from one process or contain subsets related to different states of the data producing process. Data visualization tools should deliver a clear picture of the univariate probability density distribution (PDF) for each feature. Visualization tools for PDFs typically use kernel density estimates and include both the classical histogram, as well as the modern tools like ridgeline plots, bean plots and violin plots. If density estimation parameters remain in a default setting, conventional methods pose several problems when visualizing the PDF of uniform, multimodal, skewed distributions and distributions with clipped data, For that reason, a new visualization tool called the mirrored density plot (MD plot), which is specifically designed to discover interesting structures in continuous features, is proposed. The MD plot does not require adjusting any parameters of density estimation, which is what may make the use of this plot compelling particularly to non-experts. The visualization tools in question are evaluated against statistical tests with regard to typical challenges of explorative distribution analysis. The results of the evaluation are presented using bimodal Gaussian, skewed distributions and several features with already published PDFs. In an exploratory data analysis of 12 features describing quarterly financial statements, when statistical testing poses a great difficulty, only the MD plots can identify the structure of their PDFs. In sum, the MD plot outperforms the above mentioned methods.
Conference Paper
Full-text available
The methods and possibilities of data mining for knowledge discovery in economic data are demonstrated on data of the German system of allocating tax revenues to municipalities. This system is complex and not easily understandable due to the involvement of several layers of administration and legislation. The general aim of the system is that a share of income tax revenue for a municipality (system output) should be a fixed part of the total income tax yield of each municipality (system input). Tools for the scientific exploration of empirical distributions are applied to the input and output variables of the system. The main finding is that, although the critical input variables are unimodally distributed, the output variable showed a bimodal distribution. The conclusion from this finding is that the system works in two distinct states: municipalities receive either a large share or a small share of the input. Relating these states of the system to the location of the municipality a distinct east-west gradient is found. A significantly larger percentage of East-German municipalities receive in average 5% less of the taxes share. This paper focuses on the methods and tools for such type of knowledge discovery.
Book
Full-text available
This book is published open access under a CC BY 4.0 license (for free download see http://www.springer.com/us/book/9783658205393 ) It covers aspects of unsupervised machine learning used for knowledge discovery in data science and introduces a data-driven approach to cluster analysis, the Databionic swarm(DBS). DBS consists of the 3D landscape visualization and clustering of data. The 3D landscape enables 3D printing of high-dimensional data structures.The clustering and number of clusters or an absence of cluster structure are verified by the 3D landscape at a glance. DBS is the first swarm-based technique that shows emergent properties while exploiting concepts of swarm intelligence, self-organization and the Nash equilibrium concept from game theory. It results in the elimination of a global objective function and the setting of parameters. By downloading the R package DBS can be applied to data drawn from diverse research fields and used even by non-professionals in the field of data mining. Contents • Approaches to Unsupervised Machine Learning • Methods of Visualization of High-Dimensional Data • Quality Assessments of Visualizations • Behavior-Based Systems in Data Science • Databionic Swarm (DBS) Target Groups Lecturers, students as well as non-professional users of data science, statistics, computer science, business mathematics, medicine, biology The Author Michael C. Thrun, Dipl.-Phys., successfully defended his Ph.D. in 2017 at the Philipps University of Marburg. Thrun’s advisor was the Chair of Neuroinformatics, Prof. Dr. rer. nat. Alfred G. H. Ultsch.
Conference Paper
Full-text available
Dimensionality reduction by feature extraction is commonly used to project high-dimensional data into a low-dimensional space. With the aim to create a visualization of data, only projections onto two dimensions are considered here. Self-organizing maps were chosen as the projection method, which enabled the use of the U*-Matrix as an established method to visualize data as landscapes. Owing to the availability of the 3D printing technique, this allows presenting the structure of data in an intuitive way. For this purpose, information about the height of the landscapes is used to produce a three dimensional landscape with a 3D color printer. Similarities between high-dimensional data are observed as valleys and dissimilarities as mountains or ridges. These 3D prints provide topical experts a haptic grasp of high-dimensional structures. The method will be exemplarily demonstrated on multivariate data comprising pain-related bio responses. In addition, a new R package “Umatrix” is introduced that allows the user to generate landscapes with hypsometric tints.
Chapter
Full-text available
The objective of this paper is to present a methodology for discovering comprehensible, valid, potentially innovative, and useful patterns, i.e., new knowledge, in multidimensional spatial data. Techniques from statistics, machine learning, and data mining are applied in consecutive logical steps to allow the visualization of results and the application of validation procedures at each stage. However, the approach does not end with a data cluster; rather, if such a valid cluster has been achieved, then the question is posed: “What do the clusters mean?”. Symbolic machine learning methods are employed to produce an explanation of the clusters in terms of rules employing an understandable subset of the high-dimensional data variables. This combined with canonical representatives of a cluster and consideration of the spatial distribution of the clusters lead to hypothesis on emergent data structures, that is, potential new knowledge. The approach is demonstrated on an exemplary data set of German urban districts featuring seven dimensions of land use.
Article
Full-text available
In two studies, stock market professionals (N1 = 22, N2 = 21) and laypeople (N1 = 29, N2 = 34) provided thirty-day forecasts for twenty stocks and estimated the size of their own errors as well as their own and the other group's mean errors. Both groups predicted that the errors made by professionals would be half the size of the errors made by laypeople. In reality, the errors of both groups were about the size predicted for the laypeople. Participants also estimated their ability to pick the best performing stock from two options. Both groups proved to be overconfident. Professional predictions were only successful 40% of the time, a performance below what could be expected from chance alone. Self reports and correlations between forecasts and price movements sug- gested that the professionals based their predictions on specific information of the stocks without sufficient awareness of the unreliability of this information, while the laypeople used simple heuristics based on previous price movements. During the recent worldwide stock market turbu- lence, fortunes were rapidly made and just as rapidly lost. The great majority of participants did not antici- pate either the magnitude or the pace of these events. The bewilderment has been hard to hide. This paper addresses two questions raised by more than a few people in the aftermath: "How much do stock market professionals really know about the fu- ture development of stock prices, and how much do they think they know?" We first discuss psychological findings concerning expert judgments and experts' and laypeople's confidence in their own judgments.
Article
This paper examines natural-color maps by focusing on the painted map art of Hal Shelton, the person most closely associated with developing the genre during the mid twentieth century. Advocating greater use of natural-color maps by contemporary cartographers, we discuss the advantages of natural-color maps compared to physical maps made with hypsometric tints; why natural-color maps, although admired, have remained comparatively rare; and the inadequacies of using satellite images as substitutes for natural-color maps. Seeking digital solutions, the paper then introduces techniques for designing and producing natural-color maps that are economical and within the skill range of most cartographers. The techniques, which use Adobe Photoshop software and satellite land cover data, yield maps similar in appearance to those made by Shelton, but with improved digital accuracy. Full-color illustrations show examples of Shelton’s maps and those produced by digital techniques.
Article
From the school of unemotional investing comes the classic How to Make Money in Stocks, by Wall Street analyst and publisher William ONeil. Readers new to securities will find it an excellent primer, one that relies on time-honored indicators such as quarterly earnings, market capitalization, and daily indexes. ONeils study of winning stocks stretches back to the 1960s, and he shares his insights here, describing what characterizes a growth stock, when to cut your losses (at 7 or 8 percent, no more), and how to spot a market top. The techniques in How to Make Money in Stocks are hardly revolutionary, but therein lies their strength, as ONeil claims his is a winning system in good times or bad. Investors interested in Net stocks might be disappointed--the authors first rule is that a company must show a pattern of growing profits, which disqualifies many dot coms. (Try Rule Breakers, Rule Makers for a different take.) ONeils approach to stocks is, above all, rational, and he pays little heed to market hype. Those new to investing would do well to read this book before embarking, and even more seasoned traders may find How to Make Money in Stocks a refreshing return to basics. Markets may swing bull and bear, but ONeil promises to stand firm. --Demian McLean The CANSLIM stock investing methodology outlined in How to Make Money in Stocks is a time tested method that incorporates how the equity (stock) market(s) really work -for the passive, minority, outside investor. It is a 80/20 approach (with William ONeils approach, proprietary metrics and tools you can achieve 80% of the success with 20% of the effort) and explicitly rides the coattails of the market (read and react -dont fight/argue with the market). This system *IS* designed for the individual investor and small professional investor -Im not sure it would scale for $250 million portfolios and above (but Im not sure that it wouldnt either). CANSLIM is part fundamental (here earnings growth is the primary focus), part structural, part timing/technical and part money/risk management (this part is **crucial** to investment success). The vast amount of other investing books will typically provide only a screening/selection approach, but How to Make Money in Stocks provides a comprehensive investment approach including stock selection, portfolio composition, selling criteria and money/risk management.