• Keine Ergebnisse gefunden

Visual Business Analytics of hierarchical Data

7.1. VISUAL ANALYTICS OF FREQUENT PATTERNS 103

7.2.4 VisMap Application Examples

The next section now shows, how we successfully employedVisMap in financial and service level analyis.

7.2.4 VisMap Application Examples

SLA Analysis

Service Level Agreements (SLAs) are very common in business scenarios. SLAs are service contracts which are signed between customers and suppliers to guarantee

7.2. VISMAP 113

that suppliers deliver certain goods and services to customers. Such a contract typically contains Service Level Objectives (SLOs) defining what service should be delivered with what level of quality and within what specified time period.

An important question business managers need to understand is whether their business operations are fulfilling SLOs and if not, which SLO is violated and what is the cause of the violation. Since the violation of SLO’s may result in enormous costs for service providers, it is of course an important issue to be able to visually analyze and monitor SLA’s. At HP Research Labs we successfully appliedVisMap to real world SLA data, for visual root cause analysis.

Figure 7.17: Visual Service Contract Analysis of an HTTP Order Entry SLO: VisMap allows different views on the data via Multi-Attributes Layered Drilldown and Correlation Analysis

Figures 7.17 illustrates a series of drilldowns to find the cause of an HTTP Order Entry SLO violation. The initial VisMap Layout aligns therefore all SLO’s (OrderEntry, Email) based on their violation levels per hour. To identify SLO violations, the maximum violation level per hour is mapped to color. For example, in the upper image in Figure 7.17, the analyst discovers an SLO violation (Or-derEntry) at the 7th hour. Next, the analyst drills down from the Order Entry, discovers that the cause of the violation is an Oracle database, shown in the

mid-114 CHAPTER 7. ANALYSIS OF HIERARCHICAL DATA

dle image in Figures 7.17. It’s SLO violation level is 0.231 percent. To find the cause for this violation the analyst drills down from Oracle DB, to the DB’s logged system performance metrics (availability, response time, and setup time) and finds that the cause of the violation is that system availability is low, which makes the response time long, i.e. above the allowed threshold, and thus the Order Entry contract is violated. After the system support increases the system availability, the response time is reduced and the SLO violation measurement drops. Analysts can observe those changes along the time line.

Business people want to have a set of reports to compare their yearly sales.

They want to answer questions like which region, countries, and product has the most sales and which have the least sales. VisMap generates a sequence of maps to answer their questions as illustrated in Figure 7.17

Stock Market Analysis

The financial sector is an important domain dealing with complex time dependent data sets. The visual analysis of these kinds of data is an essential issue in tech-nical finance market analysis to support asset performance analysis and decision making processes. In financial analysis, however, the most important and most common visualization techniques for time series data are chart diagram, typically line- and bar charts, since they provide an intuitive way to get insight into price fluctuations of securities and assets. These charts may be enriched by overlaying aggregate plots, e.g. moving averages. One of the most important asset price series characteristic from an analyst’s point of view is Return. RegardingReturn, analysts and investors are interested in growth rates of an asset price series within certain, often multiple, different time frames. Briefly, growth rate is defined as the ratio between the asset price at the end and the starting point of a time frame interval.

Since there are typically thousands of assets in the market separated in different categories, a performance analysis of each one over time using charts is a difficult task, since numerous charts have to be constructed and compared. Therefore we used VisMap to analyze such large numbers of assets over time. We used the Squarified Treemap approach [BHvW00] to visualize the different finance sectors, similar to Figure 7.13. However, in contrast to the technique in Figure 7.13, we now use visual aligned layouts to show the asset performance over time, and not only at a certain point in time. This makes it easy to identify correlations of asset performance over time and to identify good and bad assets. We analyzed the performance of 920 funds contained in the Lipper Bond Index. In this database, all prices were sampled on a monthly basis during March 2002 and March 2005, whereas not each asset covered the whole time frame. The database represents European and international funds composed of stock assets partitioned into several sectors. The initial layout is shown in Figure 7.18. The sectors of the layout represent the different financial sectors whereas the size of each box reflects the number of funds belonging to that sector. As shown in the Figure, most funds

7.2. VISMAP 115

Figure 7.18: VisMap using Rectangular Layouts to analyze 920 Lipper Bond Stocks per sector via aligned hierarchical layouts. It is easy to identify strong (green) and weak (red) periods for each sector / asset. The funds per sector are aligned per day, and sorted by overall performance. It is easy to see the negative impact of the SARS epidemic and the Iraq war in early 2003 to the stock market.

belong to the technology sectors. Color shows the normalized stock price per asset over the whole time span.

Different ordering functions are applied to improve the visual representation.

At first the funds are aligned according to their time stamps. It is easy to identify funds that were floated after March 2002. It is also easy to see that the SARS epidemic and the begin of the war in Iraq had a major impact on the stock market, leading to falling stock prices across all sectors. However, it is easy to see that fund of the finance and industry sector recovered much faster from these negative effects, than the technology or pharma sector. As shown in Figure 7.18 interactive tools like tool tip functionality can be used to get detailed information for certain funds. VisMap supports different visual methaphors. Suppose that the analyst is now interested to compare the average performances of all sectors compared to a certain query point (e.g. point of purchase). Therefore he selects the CircleView metaphor [KSS04a] and performs a roll-up operation, as shown in Figure 7.19. The

116 CHAPTER 7. ANALYSIS OF HIERARCHICAL DATA

Figure 7.19: VisMap using Circular Layouts to analyze Lipper Bond stocks per sector. The left figure shows the average performance of each sector.

The right figure shows the avg asset price differences between each month (e.g. point of sell) and the first month of the fund (March 2002, e.g. point of purchase) after the drill down on asset level.

left figure shows the average growth rate between the query point (March 2002, shown at the outside of the circle as white Circle Segment) and each month till March 2005. Green color indicates positive growth rates, red color negative ones.

The figure shows that, if one would had invested in the technology sector in March 2002, there would have been a good chance they might have lost money till 2005.

The precious sector however would have been a good investment. If the analyst now wants more details, he may drill down to the next deeper level, and analyze particular stocks. Suppose the analyst would be interested in short term trades, than he may change the colormap to reflect monthly growth rates, shown in the right figure. The figure shows, that even in this bad overall performance period 2002 - 2005, there were possibilities to make money with short term (monthly) trades, indicated by green areas. In this manner the analyst can adapt theVisMap layout to his demands, and even may generate line charts for single funds using linked views to get details on demand.

7.2.5 Conclusion

TheVisMap technique is a novel approach to visualize multi-attribute time series, by transform hierarchical time related data to a hierarchical structured visual map.

To speed up visual comparisons, time series are ordered and aligned according to certain criteria, such as total, average, maximum, or correlation. The experimen-tal studies show significant advantages of the VisMap technique at discovering patterns, trends, and historical problems in comparison to existing space filling layout approaches.

Chapter 8

Visual Business Analytics of