Topics: Coronavirus | AI | Data Science | Deep Learning | Machine Learning | Python | R | Statistics

KDnuggets Home » News » 2015 » Dec » Opinions, Interviews, Reports » Kanri Distance approach for translating Predictive Models to Actions ( 15:n42 )

Kanri Distance approach for translating Predictive Models to Actions


Kanri proprietary combination of patented statistical and process methods provides a uniquely powerful and insightful ability to evaluate large data sets with multiple variables.



Importance of Correlation-based Kanri Distance

The Kanri distance emphasis on correlations is explained with Figure 1, in which only two variables, height and weight are considered. In this figure, the elliptical region corresponds to the KDC reference group of healthy people, from the center of which the distances are measured on the measurement scale.  From Figure 1, it is clear that both Sam and Raj will be considered to be normal when we look at the two variables independently.  Both have their height and weight levels within the range of the rectangle (gray area). However, when the correlation between the two variables is taken into account, both men are outside of the normal or reference group (elliptical region).  This is very important aspect of the Kanri distance.  In addition, with Kanri distances we can quantitatively determine how far these individuals are outside of the reference group.  The KDC measures the degree of abnormality of individual people or products with respect to the desired reference group.  In this figure it can be clearly seen that Raj is less abnormal than Sam.  This type of information will help to provide the better treatment for each person.  Selection of the reference group is an important step in KDC as the measurements and hence decisions depend on this group.

kdc-figure1
Figure 1: KDC Emphasis on Correlations

Illustrative Case Study with Outputs

Objective:  To identify contributing performance factors for 45 branches of a major bank so that strategic decisions can be made to improve overall performance and productivity.

These 45 branches were compared with a target group of high performing branches with 10 variables that were used to judge the performance.  Examples of these variables are: service time, median income, average credit and average employee tenure etc.

Figure 2 shows the Kanri distances for the target group and the test group (45 branches).  Good separation between these two groups in terms of Kanri distances validates the scale.

kdc-figure2
Figure 2: Measurement Scale Validation

Branch Distances from the Target

From the Figure 3 we can see that Branches 10, 13, 29, 42 and 44 have higher distances from the target, and should be looked into first.  This figure also shows distances of the 45 branches under study.  It is clear that X9 is most important considering all branches followed by X8.

kdc-figure3
Figure 3: Distances and Variable Contributions for the Distances

Table 1 below represents sample output for the root cause analysis showing contributions of the all factors for the branch distances.   The output is arranged in descending order to prioritize distances with higher distances.  They help us to determine contributing factors on individual basis. Kanri Distance Calculator provides individualized driver view, with impacts for each participant.  For example in Branch 10, X4 drives 17.62%, X8 drives 13.3% and X10 drives 12.14% of the distance.  Note that we will customize this analysis case-by-case basis.  Multiple participant categories can be separately diagnosed.

kdc-table1
Table 1: Sample output of Root Cause Analysis through Contribution Ratios

After identifying key contributors, Statistical Process Control (SPC) charts will be used to monitor these variables based on the limits from the target group.


Sign Up

By subscribing you accept KDnuggets Privacy Policy