Matchmaking Performance Analysis

In the current matchmaking of UDDI, if user requests with imprecise query like

“cheap airline service”, UDDI just returns the result services for “airline service”. The amount of result services may be more than those services user really want. With our

proposed matchmaking, we still can’t confirm that our result can satisfy the need of all the requester. But we can provide more precise suited result to requesters.

The measure to investigate the performance between UDDI and our mechanism is the imprecision rate. The imprecision rate means that the rate of the amount of wrong services is compared with the total search services.

Imprecision Rate = |Suser–Smatched| / TS

Suser: the amount of services user really wants Smatched: the amount of matched services TS: the amount of total services

The value of |Ruser –Rmatched| is bigger than zero when (1) the amount of the matched services is more than the amount of services user wants or (2) the amount of the matched services is less than the amount of services user wants or (3) the matched services are different from the services user wants. TS mean the total services with which we match the suited services user requests.

Imprecision rate helps us to investigate what kind of matchmaking mechanism can provide more precise result to users.

The following tables show the imprecision rates of UDDI and our ontology-based fuzzy matchmaking framework under the different situations. In our framework, we classify tickets into the tickets for traveling and the yearly tickets. In table 4-9, the slot “Price Range”means the different expectation of price for cheapness. In table 4-10, the slots “Seat Space”and “Air Time”mean the expectation of the space of one seat and the expectation time passengers take to arrive to their destinations for comfort. In table 4-11, except for “Price Range”, we have the expectation ratio for “Price Range”. It is used to estimate the quantification type of PRUF rules, “most tickets are cheap”. In table 4-12, we evaluate the imprecision rate for fuzzy intersection operation “and”.

Table 4-9 The Imprecision Rate for “cheap airline”Query

Price Range UDDI OFMWS

Tra<=15500orYear<=19500 0.3 0.5

Tra<=14500orYear<=18500 0.5 0.3

Tra <=13500orYear<=17500 0.8 0

Tra <=12500orYear<=16500 1.0 0.2

Tra <=11500orYear<=15500 1.0 0.2

Average 0.72 0.24

Table 4-10 The Imprecision Rate for “comfortable airline”Query Comfortable factors

Seat Space Air Time

UDDI OFMWS

Tra>=2.5orYear>=3 Tra<=2orYear <=2 0.4 0.2 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 0.4 0.2 Tra>=1.5orYear >=2 Tra<=3orYear <=3 0.4 0.4

Average 0.4 0.27

Table 4-11 The Imprecision Rate for “most tickets are cheap”Query

Price Range ratio UDDI OFMWS

Tra<=15500orYear<=19500 >=0.5 0.4 0.4 Tra<=15500orYear<=19500 >=0.8 0.7 0.1 Tra<=14500orYear<=18500 >=0.5 0.7 0.1

Tra<=14500orYear<=18500 >=0.8 0.8 0

Tra <=13500orYear<=17500 >=0.5 0.8 0 Tra <=13500orYear<=17500 >=0.8 1 0.2 Tra <=12500orYear<=16500 >=0.5 0.9 0.1 Tra <=12500orYear<=16500 >=0.8 1 0.2 Tra <=11500orYear<=15500 >=0.5 1 0.2 Tra <=11500orYear<=15500 >=0.8 1 0.2

Average 0.73 0.15

Table 4-12 The Imprecision Rate for “cheap and comfortable airline”Query

Price Range Seat Space Air Time UDDI OFMWS

Tra<=15500orYear<=19500 Tra>=2.5orYear>=3 Tra<=2orYear <=2 1 0.1 Tra<=15500orYear<=19500 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 0.7 0.2 Tra<=15500orYear<=19500 Tra>=1.5orYear >=2 Tra<=3orYear <=3 0.3 0.6 Tra<=14500orYear<=18500 Tra>=2.5orYear>=3 Tra<=2orYear <=2 1 0.1 Tra<=14500orYear<=18500 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 0.8 0.1 Tra<=14500orYear<=18500 Tra>=1.5orYear >=2 Tra<=3orYear <=3 0.7 0.2 Tra <=13500orYear<=17500 Tra>=2.5orYear>=3 Tra<=2orYear <=2 1 0.1

Tra <=13500orYear<=17500 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 0.9 0 Tra <=13500orYear<=17500 Tra>=1.5orYear >=2 Tra<=3orYear <=3 0.9 0 Tra <=12500orYear<=16500 Tra>=2.5orYear>=3 Tra<=2orYear <=2 1 0.1 Tra <=12500orYear<=16500 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 0.9 0 Tra <=12500orYear<=16500 Tra>=1.5orYear >=2 Tra<=3orYear <=3 0.9 0 Tra <=11500orYear<=15500 Tra>=2.5orYear>=3 Tra<=2orYear <=2 1 0.1 Tra <=11500orYear<=15500 Tra>=2orYear >=2.5 Tra<=2.5orYear<=2.5 1 0.1 Tra <=11500orYear<=15500 Tra>=1.5orYear >=2 Tra<=3orYear <=3 1 0.1

Average 0.87 0.12

With table 4-9, 4-10, 4-11, 4-12, we can observe that the imprecision rate of our proposed framework, ontology-based fuzzy matchmaking for Web services (OFMWS), is lower than that of UDDI. That is, our framework can support vague query and return more correct results to consumers. UDDI sometimes can fulfill the demands of consumers, but it can’tsupport vague query to provide precise results.

Hence, our framework always returns more suited results to satisfied user’s request.

Otherwise, consumers don’t need to browse each data of different services to find suited services. We reduce the search space by classifying Web services with fuzzy logic.

In sum, though it has some constraints such as the limitation of the parameter of OWL-S and the limitation of dynamic classifying Web services, our proposed framework explores the hidden dimension of matchmaking of Web services, and allows users to search with vague sentences. Otherwise, it reduces the search space for users and provides more fitting Web services to users.