Fig. 4.29: Class distribution of the bamboo-tea bag process
Table. 4.6: Analysis of the bamboo-tea bag process
Mean Median Standard
deviation
Maximum Minimum
F1 0.5 0.5 0.51 1 0
F2 12.58 13 1.52 15 10
F3 1.67 1 0.79 3 1
F4 0.56 1 0.50 1 0
T 1.36 2 1.15 3 0
Table.4.7: Correlation between each attribute and the target attribute
Attribute pairs Correlation coefficient
F1 and T 0.12
F2 and T 0.15
F3 and T 0.04
F4 and T 0.23
Using MATLAB and ANFIS classifier the tea-bag dataset was analyzed. The initial step was to normalize the dataset features to an interval of 0 to 1. Then the dataset was divided into testing and training data set.
The structure of ANFIS used in this case study consisted of four inputs and a single output. The number of epochs used was 3 with and error tolerance of 0.012. The first and the fifth input functions were assigned three member each and also the other three outputs are assigned three membership functions. The total number of rules generated was 81.
0 36%
1 8%
2 39%
3 17%
Fig.4.30: Training data for bamboo tea bag
Fig. 4.31: ANFIS structure for the tea bag
Fig. 4.32: Training error in ANFIS system for tea bag
Fig. 4.33: Training vs FIS output for Tea bag
Fig. 4.34: Surface map for tea bag making process
Fig.4.35: Rules for the ANFIS tea bag process
Bamboo curtain making process: The next case study which was considered in this research was the bamboo curtain making process. The bamboo curtain making process also is common to the regions where the bamboo craft is practiced and people usually
Height Distance between circular rings
Width of the circular
frame
Fig.4.36: Bamboo Curtain stand
Table 4.8: Features of the bamboo curtain process
Feature ID Feature Name Description
F1 H1 Height of the stand (152 to
170)
F2 W1 Width of the circular frame (91
to 170)
F3 D1 Distance between the circular
rings (10 to 15)
F4 D2 Distance between the frames
(40 to 60)
F5 N1 Number of frames (1 to 5)
T(target) Location (0,1,2,3) 0=North, 1=South, 2=East, 3=
West
Fig. 4.37: Class distribution of the bamboo-curtain process Table 4.9: Analysis of the bamboo-tea bag process
Mean Median Standard
deviation
Maximum Minimum
F1 162.53 164 5.26 170 152
F2 129.14 124 25.79 170 91
F3 12.64 13 1.58 15 10
F4 50.81 51.5 5.40 60 40
F5 3.06 3 1.47 5 1
T 1.56 2 1.13 3 0
Table 4.10: Correlation between each attribute and the target attribute
Attribute pairs Correlation coefficient
F1 and T -0.18
F2 and T 0.08
F3 and T -0.14
F4 and T 0.15
F5 and T -0,26
0 26%
1 21%
2 32%
3 21%
Class distribution
0 1 2 3
Using MATLAB and ANFIS classifier the bamboo curtain stand dataset was analyzed. The initial step was to normalize the dataset features to an interval of 0 to 1. Then the dataset was divided into testing and training data set.
The structure of ANFIS used in this case study consisted of four inputs and a single output. The number of epochs used was 3 with and error tolerance of 0.00001. The first and the fifth input functions were assigned three member each and also the other three outputs are assigned three membership functions. The total number of rules generated was 243.
Fig. 4.38: Training data for bamboo curtain stand
Fig. 4.39: ANFIS structure for the bamboo curtain stand
Fig. 4.40: Training error in ANFIS system for Bamboo curtain stand
Fig. 4.41: Training vs FIS output for bamboo curtain stand
Fig. 4.42: Rules generated for the ANFIS system of bamboo curtain stand
Bamboo basket making process: The ANFIS model was also develop for the bamboo basket making process. The basket making process involves many designs, only simple netting pattern was used in this study.
Height
radius Gap between
the bamboo strips
Fig. 4.43: Bamboo basket
Table 4.11: Features of the bamboo basket process
Feature ID Feature Name Description
F1 H1 Height of the basket (10 to 50)
F2 W1 Width of the basket (10 to 50)
F3 D1 Gap between the bamboo strips
(1 to 3)
F4 R1 Radius of the corners (3 to 10)
F5 N1 Number of colors used (1 to 5)
T(target) Location (0,1,2,3) 0=North, 1=South, 2=East, 3=
West
Fig. 4.44: Class distribution of the bamboo-basket process
The class distribution is as illustrated in the Fig. 4.44. As the target group is divided into four regions of the country so the craft was also chosen from the craftsman from the four different parts of the country.
Table. 4.12: Analysis of the bamboo-tea bag process
Mean Median Standard
deviation
Maximum Minimum
F1 32.08 33 13.15 50 10
F2 28.53 26 11.25 50 10
F3 1.94 2 0.79 3 1
F4 6.89 7 2.50 10 3
F5 2.94 3 1.43 5 1
T 1.22 1 0.98 3 0
Table 4.13: Correlation between each attribute and the target attribute
Attribute pairs Correlation coefficient
F1 and T -0.02
F2 and T 0.14
F3 and T -0.35
F4 and T -0.33
F5 and T -0.12
0 26%
1 21%
2 32%
3 21%
Class distribution
0 1 2 3
Using MATLAB and ANFIS classifier the bamboo basket dataset was analyzed. The initial step was to normalize the dataset features to an interval of 0 to 1. Then the dataset was divided into testing and training data set.
The structure of ANFIS used in this case study consisted of four inputs and a single output. The number of epochs used was 3 with and error tolerance of 0.00001. The first and the fifth input functions were assigned three member each and also the other three outputs are assigned three membership functions. The total number of rules generated was 243.
Fig.4.45: Training data loaded for bamboo basket process
Fig. 4.46: ANFIS structure for the bamboo basket
Fig. 4.47: Training error for Bamboo basket data
Fig. 4.48: Training Vs FIS Output data for bamboo basket process
Fig. 4.49: Rules generated for the ANFIS system for bamboo basket process.
4.4 FIS system for all the craft studies:
The FIS system for the next 4 experimental case products was also build in a similar way as described above. The methodology proposed was followed and research questions RQ1 and RQ2 were satisfied so far in this chapter. This is illustrated in the Fig. 4.50 below:
Sl.
No .
Operation performed Lessons learnt Attributes captured
1 Craftsmen starting with bamboo
Knowledge of physical and
material behavior of bamboo
Material Property
2 Cutting bamboo to
approximate sizes
Understanding of different sizes of bamboo
Sizes of bamboo
3
Building up the form of the basket by stitching the primary and secondary strips
Understand and mold the preferred shape
Stitching shapeor curve and gap
between the strips
4
Unconsciously makes error in the craft which givesauniqueshapeof thebasket
Accounts the human error and tolerances
Final shape of the basket/ human touch Sl.
No .
Operation performed Lessons learnt Attributes captured
1 Craftsmen starting with bamboo
Knowledge of physical and
material behavior of bamboo
Material Property
2 Cutting bamboo to
approximate sizes
Understanding of different sizes of bamboo
Sizes of bamboo
3
Building up the form of the basket by stitching the primary and secondary strips
Understand and mold the preferred shape
Stitching shapeor curve and gap
between the strips
4
Unconsciously makes error in the craft which givesauniqueshapeof thebasket
Accounts the human error and tolerances
Final shape of the basket/ human touch Sl.
No .
Operation performed Lessons learnt Attributes captured
1 Craftsmen starting with bamboo
Knowledge of physical and
material behavior of bamboo
Material Property
2 Cutting bamboo to
approximate sizes
Understanding of different sizes of bamboo
Sizes of bamboo
3
Building up the form of the basket by stitching the primary and secondary strips
Understand and mold the preferred shape
Stitching shapeor curve and gap
between the strips
4
Unconsciously makes error in the craft which givesauniqueshapeof thebasket
Accounts the human error and tolerances
Final shape of the bask et/ human touch
If then rules Products
Explicit Component
Tacit/implicit Component
Extraction Extraction
Exact values/
clearly defined
Repeated practices/Fuzzy
Transmission Transmission
Craftsman
Tools developed Create Machine understandable Code/
engine
Validation
Success Finish
No Yes
Inputs
Fig. 4.50: Quick recap of the methodology followed in the research
The FIS system for all the models were developed by the inputs of the crafts and the shape and size of the crafts taken.
4.5 Summary of the chapter
In this chapter the detailed analysis of the craft objects chosen in chapter-3 is illustrated. Taking the case of Diya making, the fuzzy inference system was built and all other fuzzy inference system of the other craft cases were also built. To complete the understanding of the whole craft product transfer learning was applied to the craft cases which was chosen in the chapter-3. The next chapter test the validity of the fuzzy systems and the transfer learning model.