Page 38 - Fister jr., Iztok, and Andrej Brodnik (eds.). StuCoSReC. Proceedings of the 2018 5th Student Computer Science Research Conference. Koper: University of Primorska Press, 2018
P. 38
Table 3. Classification Rules found by the PART algorithm
Classification rules Conf. Cov.
{ Ram = 16GB, Memory = SSD, ScreenResolution = Full HD } ==> { Price_euros= High } 80% 36
{ Ram = 16GB, Memory = SSD } ==> { Price_euros= High } 94% 49
{ Ram = 16GB, Inches = Gaming } ==> { Price_euros= High } 94% 49
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i3, ScreenResolution = 1366x768 } 98% 52
==> { Price_euros=Low }
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i5, Memory = HDD } 74% 48
==> {Price_euros=Low}
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Celeron } ==> {Price_euros=Low} 100% 48
{ Cpu = Intel Core i3, Memory = HDD } ==> { Price_euros=Low } 98% 47
{ Ram = 16GB } ==> { Price_euros= High } 70% 35
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i5 } ==> { Price_euros= Low } 49% 17
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Other } ==> { Price_euros= Low } 100% 30
{ Ram = 4GB, ProductTypeName = Notebook } ==> { Price_euros= Low } 97% 58
{ Cpu = Intel Core i7, ProductTypeName = Ultrabook, Inches = Standard } ==> { Price_euros= High } 70% 31
{ Cpu = Intel Core i7, ProductTypeName = Gaming, Inches = Standard } 61% 33
==> { Price_euros= Mid-Range }
{ Cpu = Intel Core i7, Inches = Standard, Memory = SSD } ==> { Price_euros= Mid-Range } 51% 33
{ Cpu = Intel Core i7, Inches = Small } ==> { Price_euros=High } 71% 41
{ Cpu = Intel Celeron } ==> { Price_euros= Low } 100% 40
{ ProductTypeName = Gaming, Cpu = Intel Core i7} ==> { Price_euros= High } 76% 29
{ Memory = SSD + HDD } ==> { Price_euros= Mid-Range } 58% 42
{ Cpu = Intel Core i7} ==> { Price_euros= Mid-Range } 49% 23
{ Cpu = Intel Core i5, Weight = Light, Company = Dell } ==> { Price_euros= Mid-Range } 49% 18
{ Cpu = Intel Core i5, Weight = Light, ProductTypeName = Ultrabook } ==> {Price_euros= Mid-Range } 50% 21
{ Cpu = Intel Core i5, Memory = SSD, Weight = Light } ==> {Price_euros= Mid-Range } 49% 30
{ Cpu = Intel Core i5, Gpu = Intel } ==> {Price_euros= Mid-Range } 48% 38
5. CONCLUSIONS AND FUTURE WORK [3] Cendrowska, J. (1987). “PRISM: An algorithm for inducing
modular rules”. International Journal of Man Machine
Overall, it can be observed from the experimental results (Apriori Studies. volume 27, no. 4, pp. 349-370.
algorithm) that the features of the cheaper laptops (rules in
Table 2 with the right-hand side Price_euros=Low) were inches: [4] Duda, R., and Hart, P. (1973). “Pattern Classification and
Standard or Small, CPU: Intel Core i3 or Celeron, RAM: 4GB, Scene Analysis”. John Wiley & Sons.
Weight: Light or Medium, Memory: HDD, GPU: Intel, Operative
System: Windows 10 and Screen Resolution: 1366x768. [5] Frank, E., and Witten, I. (1998). “Generating accurate rule
sets without global optimization”. Proceedings of the
The expensive laptops (rules in Table 2 with the right-hand side Fifteenth International Conference on Machine Learning,
Price_euros=High) were mostly Gaming laptops with SSD or Morgan Kaufmann, San Francisco, pp. 144-151.
SSD+HDD memory, 16GB RAM, Intel Core i7 CPU, Nvidia
GPU, Heavy weight and Full HD Screen Resolution. [6] Fürnkranz, J. (1996). “Separate-and-conquer rule learning”.
Technical Report TR-96-25, Austrian Research Institute for
The results on the chosen dataset show that the Apriori algorithm Artificial Intelligence, Vienna.
generates more accurate individual classification rules with higher
coverage/accuracy than the PART algorithm. However, class [7] Fürnkranz, J., and Widmer, G. (1994). “Incremental reduced
association rules may be highly overlapping which can affect the error pruning”. Proceedings of the 11th Annual Conference
overall accuracy of the classifier. on Machine Learning, Morgan Kaufmann, pp. 70-77.
In future work we shall check the overall accuracy and coverage [8] John, G.H., and Langley, P. (1995). “Estimating Continuous
of “complete” classifiers generated by class association rules Distributions in Bayesian Classifiers”. Proceedings of the
algorithm, broaden our comparison to more (diverse) datasets and Eleventh Conference on Uncertainty in Artificial
compare also the learning times of both algorithms. Intelligence. Morgan Kaufmann, San Mateo. pp. 338-345.
6. REFERENCES [9] Liu, B., Hsu, W., and Ma, Y. (1998). “Integrating
Classification and Association Rule Mining”. Singapore.
[1] Agrawal, R., Amielinski, T., and Swami, A. (1993). “Mining
association rule between sets of items in large databases”. In [10] Patel, R., Vala, J., and Patel, K. (2014). “Comparative Study
Proceeding of the ACM SIGMOD International Conference on Associative Classification techniques”.
on Management of Data, pp. 207-216.
[11] Quinlan, J. R. (1993). “C4.5: Programs for Machine
[2] Agrawal, R., and Srikant, R. (1994). “Fast algorithms for Learning”. San Mateo, Morgan Kaufmann.
mining association rules”. Research Report CA 95120, IBM
Almaden Research Center, San Jose, California. [12] Quinlan, J. R. (2008). “Data mining tools see5 and c5”.
[13] https://www.kaggle.com/ionaskel/laptop-prices/home
(accessed on 17.8.2018).
[14] http://software.ucv.ro/~cmihaescu/ro/teaching/AIR/docs/Lab
8-Apriori (accessed on 17.8.2018).
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference 38
Ljubljana, Slovenia, 9 October
Classification rules Conf. Cov.
{ Ram = 16GB, Memory = SSD, ScreenResolution = Full HD } ==> { Price_euros= High } 80% 36
{ Ram = 16GB, Memory = SSD } ==> { Price_euros= High } 94% 49
{ Ram = 16GB, Inches = Gaming } ==> { Price_euros= High } 94% 49
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i3, ScreenResolution = 1366x768 } 98% 52
==> { Price_euros=Low }
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i5, Memory = HDD } 74% 48
==> {Price_euros=Low}
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Celeron } ==> {Price_euros=Low} 100% 48
{ Cpu = Intel Core i3, Memory = HDD } ==> { Price_euros=Low } 98% 47
{ Ram = 16GB } ==> { Price_euros= High } 70% 35
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Core i5 } ==> { Price_euros= Low } 49% 17
{ Ram = 4GB, ProductTypeName = Notebook, Cpu = Intel Other } ==> { Price_euros= Low } 100% 30
{ Ram = 4GB, ProductTypeName = Notebook } ==> { Price_euros= Low } 97% 58
{ Cpu = Intel Core i7, ProductTypeName = Ultrabook, Inches = Standard } ==> { Price_euros= High } 70% 31
{ Cpu = Intel Core i7, ProductTypeName = Gaming, Inches = Standard } 61% 33
==> { Price_euros= Mid-Range }
{ Cpu = Intel Core i7, Inches = Standard, Memory = SSD } ==> { Price_euros= Mid-Range } 51% 33
{ Cpu = Intel Core i7, Inches = Small } ==> { Price_euros=High } 71% 41
{ Cpu = Intel Celeron } ==> { Price_euros= Low } 100% 40
{ ProductTypeName = Gaming, Cpu = Intel Core i7} ==> { Price_euros= High } 76% 29
{ Memory = SSD + HDD } ==> { Price_euros= Mid-Range } 58% 42
{ Cpu = Intel Core i7} ==> { Price_euros= Mid-Range } 49% 23
{ Cpu = Intel Core i5, Weight = Light, Company = Dell } ==> { Price_euros= Mid-Range } 49% 18
{ Cpu = Intel Core i5, Weight = Light, ProductTypeName = Ultrabook } ==> {Price_euros= Mid-Range } 50% 21
{ Cpu = Intel Core i5, Memory = SSD, Weight = Light } ==> {Price_euros= Mid-Range } 49% 30
{ Cpu = Intel Core i5, Gpu = Intel } ==> {Price_euros= Mid-Range } 48% 38
5. CONCLUSIONS AND FUTURE WORK [3] Cendrowska, J. (1987). “PRISM: An algorithm for inducing
modular rules”. International Journal of Man Machine
Overall, it can be observed from the experimental results (Apriori Studies. volume 27, no. 4, pp. 349-370.
algorithm) that the features of the cheaper laptops (rules in
Table 2 with the right-hand side Price_euros=Low) were inches: [4] Duda, R., and Hart, P. (1973). “Pattern Classification and
Standard or Small, CPU: Intel Core i3 or Celeron, RAM: 4GB, Scene Analysis”. John Wiley & Sons.
Weight: Light or Medium, Memory: HDD, GPU: Intel, Operative
System: Windows 10 and Screen Resolution: 1366x768. [5] Frank, E., and Witten, I. (1998). “Generating accurate rule
sets without global optimization”. Proceedings of the
The expensive laptops (rules in Table 2 with the right-hand side Fifteenth International Conference on Machine Learning,
Price_euros=High) were mostly Gaming laptops with SSD or Morgan Kaufmann, San Francisco, pp. 144-151.
SSD+HDD memory, 16GB RAM, Intel Core i7 CPU, Nvidia
GPU, Heavy weight and Full HD Screen Resolution. [6] Fürnkranz, J. (1996). “Separate-and-conquer rule learning”.
Technical Report TR-96-25, Austrian Research Institute for
The results on the chosen dataset show that the Apriori algorithm Artificial Intelligence, Vienna.
generates more accurate individual classification rules with higher
coverage/accuracy than the PART algorithm. However, class [7] Fürnkranz, J., and Widmer, G. (1994). “Incremental reduced
association rules may be highly overlapping which can affect the error pruning”. Proceedings of the 11th Annual Conference
overall accuracy of the classifier. on Machine Learning, Morgan Kaufmann, pp. 70-77.
In future work we shall check the overall accuracy and coverage [8] John, G.H., and Langley, P. (1995). “Estimating Continuous
of “complete” classifiers generated by class association rules Distributions in Bayesian Classifiers”. Proceedings of the
algorithm, broaden our comparison to more (diverse) datasets and Eleventh Conference on Uncertainty in Artificial
compare also the learning times of both algorithms. Intelligence. Morgan Kaufmann, San Mateo. pp. 338-345.
6. REFERENCES [9] Liu, B., Hsu, W., and Ma, Y. (1998). “Integrating
Classification and Association Rule Mining”. Singapore.
[1] Agrawal, R., Amielinski, T., and Swami, A. (1993). “Mining
association rule between sets of items in large databases”. In [10] Patel, R., Vala, J., and Patel, K. (2014). “Comparative Study
Proceeding of the ACM SIGMOD International Conference on Associative Classification techniques”.
on Management of Data, pp. 207-216.
[11] Quinlan, J. R. (1993). “C4.5: Programs for Machine
[2] Agrawal, R., and Srikant, R. (1994). “Fast algorithms for Learning”. San Mateo, Morgan Kaufmann.
mining association rules”. Research Report CA 95120, IBM
Almaden Research Center, San Jose, California. [12] Quinlan, J. R. (2008). “Data mining tools see5 and c5”.
[13] https://www.kaggle.com/ionaskel/laptop-prices/home
(accessed on 17.8.2018).
[14] http://software.ucv.ro/~cmihaescu/ro/teaching/AIR/docs/Lab
8-Apriori (accessed on 17.8.2018).
StuCoSReC Proceedings of the 2018 5th Student Computer Science Research Conference 38
Ljubljana, Slovenia, 9 October
