3.3.1. Principal Component Analysis
To screen the key quality indexes of pork under the methods of boiling, scalding, and roasting, PCA was conducted for each quality index. The results of PCA and the weight of each index under the three cooking methods are shown in
Table 2,
Table 3 and
Table 4.
As can be seen from
Table 2, under the boiling method, the quality indexes of different pork muscles were separately analyzed for the four principal components, each of which had a characteristic root greater than 1, and the cumulative variance contribution rate reached 81%. The three-dimensional(3D) loading diagram under boiling method can be seen in
Figure S1. The absolute weights of flesh color, elasticity, marbling,
L*,
b*, drip loss, shearing force, cholesterol, inosinic acid, and UFA were lower than others, so they were not considered in the selection of indicators.
As indicated in
Table 3, under the scalding method, the quality indexes of different pork muscles were analyzed for the four principal components, each of which had a characteristic root greater than 1, and the cumulative variance contribution rate reached 81%. The 3D loading diagram under scalding method can be seen from
Figure S2. The absolute weights of flesh color, elasticity, marbling,
L*, drip loss, shearing force, cholesterol, inosinic acid, and UFA were lower than others, so they were not considered in the selection of indicators.
As can be seen from
Table 4, the quality indexes of different pork muscles were analyzed for five principal components under the roasting method, each of which had a characteristic root greater than 1, and the cumulative variance contribution rate reached 82%. The 3D loading diagram under roasting method can be seen from
Figure S3. The absolute weights of
L*,
a*,
b*, fat, cholesterol, and UFA were lower than others, so they were not considered when indexes were selected.
3.3.2. Cluster Analysis
To determine key quality indicators of pork under different cooking methods, cluster analysis was performed for all the quality indicators and the results were shown in
Figure 2.
Under the boiling method, the dendrogram of the quality indexes was shown in
Figure 2a. According to the clustering distance, tenderness, chewiness, juiciness, odor, flavor, fat content,
a*,
b*, and UFA were clustered into a category, while marbling, cholesterol, flesh color, and elasticity were clustered into another. The protein content, flavored amino acid, drip loss, inosinic acid, and
L* were classified into a category, while the shearing force was classified into another category alone.
Under the scalding method, the dendrogram of the quality indexes was shown in
Figure 2b. According to the clustering distance, tenderness, chewiness, juiciness, flavor, fat content, odor,
a*, and
b* were grouped into a category, UFA was grouped into a category alone, and marbling, cholesterol, flesh color, and elasticity were grouped into a category. The protein content, flavored amino acid, drip loss, inosinic acid, and
L* were classified into a category, while the shearing force was classified into a category alone.
Under the roasting method, the dendrogram of the quality indexes was shown in
Figure 2c. According to the clustering distance, tenderness, juiciness, chewiness, cooked flesh color, fat content, odor, flavor,
a*, and
b* were grouped into a category, UFA was grouped into a category alone, and marbling, cholesterol, flesh color, and elasticity were grouped into a category. Protein, flavored amino acid, drip loss, inosinic acid, and
L* were classified into a category, while the shearing force was classified into another category alone.
3.3.3. Correlation Analysis
To determine key quality indicators of pork under different cooking methods, correlation analysis was carried out for each quality index and the results were shown in
Figure 3.
As shown in
Figure 3a, under the boiling method, tenderness was significantly positively correlated with chewiness and juiciness, with correlation coefficients of 0.967 and 0.868, respectively (
p < 0.01). Tenderness is mainly related to myofibrillar protein and connective tissue protein [
41]; Sasaki et al. [
42] found that tenderness was related to chewiness, juiciness and hardness through experiments. The fat was significantly negatively correlated with the content of flavored amino acid and protein, and the correlation coefficients were −0.920 and −0.904, respectively (
p < 0.01). Amino acid are the building blocks of proteins. Proteins are broken down to produce amino acids. The content of flavored amino acid was significantly positively correlated with that of protein, and the correlation coefficient was 0.925 (
p < 0.01).
As shown in
Figure 3b, under the scalding method, tenderness was significantly positively correlated with chewiness and juiciness, with correlation coefficients of 0.957 and 0.923, respectively (
p < 0.01). The content of fat was significantly negatively correlated with the content of flavored amino acid and protein, with the correlation coefficients of −0.920 and −0.904, respectively (
p < 0.01). The content of flavored amino acid was significantly positively correlated with that of protein, with the correlation coefficient of 0.925 (
p < 0.01).
It can be seen from
Figure 3c that, under the roasting method, flavor and odor were significantly positively correlated (r = 0.839,
p < 0.01). Tenderness was significantly positively correlated with chewiness (r = 0.921,
p < 0.01) and juiciness (r = 0.942,
p < 0.01). With the loss of water, myoglobin and soluble pigment in the muscle are lost [
43], so drip loss was positively correlated flesh color (r = −0.769,
p < 0.01). Drip loss was also positively correlated with inosinic acid (r = 0.899,
p < 0.01) and protein (r = 0.832,
p < 0.01).
3.3.4. Screening of Key Quality Indicators
Key quality indicators were screened according to the weight, correlation coefficient, and coefficient of variation of each quality indicator obtained by PCA.
When key quality indexes were screened under the boiling method, the first type of index included tenderness, chewiness, juiciness, odor, flavor, fat,
a*,
b*, and UFA. As
b* and UFA had a small weight, they were not considered. Tenderness was positively correlated with chewiness and juiciness, with the correlation coefficients of 0.967 and 0.868 (
p < 0.01), coefficient of variation of 18.14%, and weight of 0.2250. Therefore, tenderness was selected as a key index. The weights of odor, flavor, and fat were 0.2162, 0.2298, and 0.1820, respectively, with the coefficients of variation of 11.44%, 14.14%, and 120.57%. Due to the large weights and coefficients of variation of odor, flavor, and fat, the three indexes were also selected. The second type of indicator was marbling, cholesterol, flesh color, and elasticity. Because the weights of the four indicators were relatively low, the second type of indicator was not selected. The third type was the protein content, flavored amino acid, drip loss, inosinic acid, and
L*. Due to the low weights of drip loss, inosinic acid, and
L*, these indicators were not selected. The weights of protein and flavored amino acid were −0.1517 and −0.1600, their coefficients of variation were 20.26% and 25.71%, and the correlation between them was 0.925 (
p < 0.01). Therefore, flavored amino acid was selected as the third representative index. The fourth type of index was the shearing force, whose weight was too low (−0.0996) so it was not considered. Since the fat was significantly correlated with flavored amino acid, with a correlation coefficient of −0.920, the fat with a higher weight and coefficient of variation was selected to replace flavored amino acid. To sum up,
a*, fat, odor, tenderness, and flavor were selected as the key quality indexes under the boiling method, and their weights were reshown in
Table 5.
In the screening of key quality indexes for scalded pork in a hot pot, the first type of index included tenderness, chewiness, juiciness, flavor, fat content, odor,
a*, and
b*. Because tenderness was positively correlated with chewiness and juiciness, with correlation coefficients of 0.957 and 0.923 (
p < 0.01), weight of 0.2086, and coefficient of variation of 15.29%, tenderness was selected as a key index. The correlation between
b* and
a* was 0.777 (
p < 0.01), the weight of
a* was 0.1479, and its coefficient of variation was 27.36%. Therefore,
a* was selected as a key index. The correlation between odor and flavor was 0.840 (
p < 0.01), and the weights of odor and flavor were 0.2073 and 0.2243, both of which were large. Therefore, odor and flavor were selected as key indicators. In short, tenderness, odor, flavor, fat, and
a* were selected to represent the first category. The second type of index was UFA, and because of its low weight of 0.0692 this type of index was deleted. The third category of indicators was marbling, cholesterol, flesh color, and elasticity, and because these four indicators had low weights this category of indicators was not chosen either. The fourth type of index was the protein content, flavored amino acid, drip loss, inosinic acid, and
L*. As the weights of
L*, drip loss, and inosinic acid were relatively low, 0.0516, −0.1147, and −0.1065, respectively, these three indexes were deleted. In this category, the weights of protein and flavored amino acid were −0.1452 and −0.1481, their coefficients of variation were 20.26% and 25.71%, and the correlation between them was 0.925 (
p < 0.01). Therefore, flavored amino acid was selected as the fourth representative index. The fifth index was the shearing force, whose weight was too low at −0.0691, so it was not considered. Since the fat was significantly correlated with flavored amino acid, with a correlation coefficient of −0.920, the fat with a larger weight and coefficient of variation was selected to replace flavored amino acid. In summary,
a*, fat, odor, tenderness, and flavor were selected as the key evaluation indexes under the scalding method, as shown in
Table 5.
When key quality indexes under the roasting method were screened, the first type of index was tenderness, juiciness, chewiness, cooked flesh color, fat content, odor, flavor,
a*, and
b*. Due to the low weights of
a*,
b*, and fat, these indexes were deleted. Tenderness was positively correlated with chewiness (r = 0.921,
p < 0.01) and juiciness (r = 0.942,
p < 0.01), its weight was 0.1620, and its coefficient of variation was 13.94%. Therefore, tenderness was selected as a key index. There was a significant positive correlation between flavor and odor (r = 0.839,
p < 0.01), the weight of flavor was 0.1379, and its coefficient of variation was 11.23%. Therefore, flavor was selected as a key index. In short, tenderness, cooked flesh color, and flavor were selected to represent the first category of indicators. The second type of index was UFA, and this type of index was deleted due to its low weight of 0.0692. The third type was marbling, cholesterol, flesh color, and elasticity. The weight of cholesterol was low, at 0.0963, so this index was excluded. The weight of elasticity was 0.1408, and its coefficient of variation was 11.54%. The weight of marbling was 0.1395, and its coefficient of variation was 14.91%. The weight of flesh color was 0.1653, and its coefficient of variation was 15.3%. Therefore, elasticity, marbling, and flesh color were selected to represent the third type. The fourth type of index was the protein content, flavored amino acid, drip loss, inosinic acid, and
L*. Since
L* had a low weight of −0.0798, this index was not selected. Drip loss was positively correlated with inosinic acid (r = 0.899,
p < 0.01) and the protein content (r = 0.832,
p < 0.01). Therefore, drip loss with a larger coefficient of variation was selected to represent the fourth type of index. The fifth type was the shearing force, whose weight, of 0.0111, was too low, so it was not considered. There was a significant negative correlation between cooked flesh color and drip loss (r = −0.769,
p < 0.01). In order to avoid a negative comprehensive score, cooked flesh color was chosen to represent drip loss. Therefore, six indexes including flavor, marbling, elasticity, cooked flesh color, tenderness, and flesh color were selected as the key indexes under the roasting method, and the weights were reshown in
Table 5.
3.3.5. Model Construction and Suitability Evaluation
According to the selected key quality indicators and weights, an evaluation model for the cooking suitability of different muscles of meat was constructed. According to the comprehensive quality evaluation scores and K-means clustering analysis, the suitability of the four muscles of pork under the three cooking methods was obtained and the results were shown in
Table 6. The score of comprehensive quality evaluation was set as
Y. Under the boiling method,
Y greater than 0.7131 was the most suitable for boiling muscles, 0.3058~0.7131 was relatively suitable for boiling muscles, and less than 0.3058 was unsuitable for boiling muscles. Under the scalding method,
Y greater than 0.647 was the most suitable, 0.251~0.647 was relatively suitable, and less than 0.251 was unsuitable. Under the roasting method,
Y greater than 0.4469 was the most suitable for roasting, 0.2344~0.4469 was relatively suitable for roasting, and
Y less than 0.2344 was not suitable for roasting.
The evaluation model under the boiling method was:
X1~
X5 represent
a*, fat, odor, tenderness, and flavor, respectively. After the determination results of the key quality indicators for boiling were normalization and standardized, they were substituted into the comprehensive quality model to calculate the scores, as shown in
Table 6. The results showed that one of the most suitable muscles for boiling was pork belly.
The evaluation model under the scaling method was:
X1~
X5 represent
a*, fat, odor, tenderness, and flavor, respectively. The measured results of the key quality indicators for scalding were normalization and standardized into the comprehensive quality model to calculate the scores, as shown in
Table 6. The most suitable muscle for scalding was pork belly.
The evaluation model under the roasting method was:
X1~
X6 represent flavor, marbling, elasticity, cooked flesh color, tenderness, and flesh color, respectively. The measured results of the key quality indicators of roasting were normalized and standardized, and then substituted into the comprehensive quality model to calculate the scores, as shown in
Table 6. The most suitable muscles for roasting were pork belly and shoulder.