Notes: WWW04-QoS computation and processing ing in Dynamic Web Service Selection

Y Liu, Ah Ngu, LZ Zeng. qoS computation and processing ing in Dynamic Web Service selection, WWW alt. '04: Proc. of 13th Int. WWW Conf. on alternate track papers & posters, 2004, pp. 66-73.
The first author, Yutu Liu, and dblp, shows this article. More information about this author is not found in other places (including EI). Zeng LZ is the third author of this article.

1. this article proposes a set of "open, fair and dynamic QoS computation model", which is used to evaluate the QoS of a single web service, the Web service composition scenario is not considered (the problem of making ws composite achieve the optimal global QoS is not solved ).

2. in general, execution price refers to the price to be paid to call this WS, but it has nothing to do with this ws; but the introduction of execution price in (s2.1, it seems that the execution price is treated as the price that some ws need to pay during execution (such as purchasing a ticket or paying a telephone fee ). I think (s2.1) is inappropriate.

3. Innovation in this article (the three points below are the innovation points of the author's Overview) (S1)

(1) extensible QoS model
Including generic and domain specific criteria.
I think this "extensible" is a bit virtual, and there are many QoS types in WS. If specific domain is involved, it is even more numerous. this article introduces three generic quality criteria (Execution price, execution duration, reputation) and three business related criteria (transaction, compensation rate, penalty rate ), the following uses QoS as an example to describe the QoS evaluation method.
The evaluation method has little to do with the specific QoS, so QoS is called "extensible ".
(2) When Evaluating QoS, it can be based on the user's preference
This is embodied in the weighted addition of formula (10). This "weight" reflects the user's preference.
(3) Fair & Open QoS Computation
QoS attribute information can be obtained from three channels
A. By provider publish (for example, price)
B. user monitoring (such as execution duration)
C. User-based feedback (such as reputation)

These innovations are common at present. They can be published on WWW, which may be the reason why the author proposed it earlier.

4. ws QoS Evaluation Method: quadratic regularization of QoS, weighted addition. (s2.2)

This section is not technically difficult, but it feels a little complicated.

The first regularization is to de-unitization and convert "decreasing measure" to "Increasing measure" (the higher the value, the more valuable ).
Here we use a different method than [Zeng lz03]
1) set a maximum value for each attribute.
2) the average value is used as the benchmark for regularization. When the QoS value is the average value, the value after regularization is 1. The higher the value after regularization, the more valuable it is.
Question: (s2.2) in example1, after the first normalization, the data given in this article is
(1.3, 1.0, 0.462, 0.769, 0.64, 0.7, 0.8894,
0.8134, 1.0, 1.538, 1.23, 3.0, 1.75, 1.111)
What I calculated is:
(1.3, 1.0, 0.462, 0.769, 0.60, 0.7, 0.9984,
0.8125, 1.0, 1.538, 1.23, 3.0, 1.75, 1.111)
Three data types are different: 0.64-> 0.60, 0.8894-> 0.9984, 0.8134-> 0.8125

Wondering, I should be right. Shouldn't such a low-level Meeting paper be wrong?

The second regularization is required because the author mentioned the concept of "Quality Group" and combined some attributes into one attribute through matrix multiplication, then, normalize these attributes (the method is the same as the method used in the first regularization), and then obtain the QoS score of WS by adding weights. I am wondering, after two regularization, the meaning of the final attributes and their values is not intuitive. How practical are these new attributes? Is it really necessary to introduce the "Quality Group" concept?