000			04009nam a22002777a 4500
999			_c83252 _d83252
003			20221105110027.0
005			20230317104408.0
008			221105b ||||| |||| 00| 0 eng d
041			_aeng
082			_223 _a571.5380285
100	1		_aMalangsa, Rhoderick D. _qDargantes _eauthor
245			_aContamination grading system for abaca tissue culture (in vitro) using Naive Bayes and K-NN algorithms / _cby Rhoderick D. Manglasa
300			_a101 [6 unnumbered] leaves: _bcolor illustrations; _c28 cm
336			_atext _btxt _2rdacontent
337			_aunmediated _bn _2rdamedia
338			_avolume _bnc _2rdacarrier
502			_aThesis (DIT) -- Cebu Institute of Technology University, College of Computer Studies, _dMarch 2017
504			_aIncludes bibliographical references.
520	3		_aThis study aims to strengthen the abaca rehabilitation program of Department of Agriculture through the use of intelligent systems for grading abaca contamination so that rapid production of high quality contaminated-free abaca specimen and uniform planting materials can be achieved. By minimizing human interaction with the abaca tissue culture in vitro specimen it can augment the current manual contamination grading in the tissue culture laboratory. Moreover, this study demonstrated the capability of the Naïve Bayes and K-NN algorithms in detecting the contaminants of abaca tissue culture in vitro specimen. The chosen neighbours for KNN were K=3 and K=7. In phase one, capturing specimen’s vision data using a camera, the distance of the camera from specimen was 32cm, elevation was 23cm, and angle was 95 degrees. Phase two, feature extraction of RGB mean values and masking was applied on specified threshold value that is computed for pixels. The red, green, blue, whitish and brownish range feature values were used for building a classification model. Based from the classification model, the Naïve Bayes algorithm can predict the likelihood of contaminations in in vitro specimen. On the other hand, KNN algorithms can predict contamination based on the selected neighbours using Euclidean distance. Furthermore, the performance of the two classifiers was evaluated based on the overall Accuracy, Precision, and Recall. Based on the findings, the Naïve Bayes classifier has recall of 97% in contaminated specimen in binary class classification and has the higher over-all accuracy of 92% than that of K=3 and K=7, 72% and 62% respectively. It was also found out in multi-class classification that Naïve Bayes performs well in Precision of 92% for healthy specimen while KNN performs well in Recall of 88% in heavily contaminated specimen. The average overall Accuracy of the Naïve Bayes was 76%, while for KNN with K=3 was 68%, KNN with K=7 was 58%. Increasing the number of neighbour for KNN algorithm in this study did not guarantee a good classification result. The study indicated that Naïve Bayes has good potential for identifying contamination grading accurately than the K-Nearest Neighbour in abaca tissue culture laboratory. The categorical form of dataset helped the Naïve Bayes in achieving a good accuracy in this study. Some factors to be considered during the training phase like the distance of the camera to the specimens, light illumination, angle of the camera, and processing power. The system cannot match the precision and recall of the human eye, but the speed and the cost can be easily overcome. It is recommended to integrate robotic arms to automatically pick classified contaminated specimen in the laboratory. Create a mobile application for ease of monitoring and notification of contaminated in vitro specimens. Lastly, for future work apply fuzzy logic on the contamination grading since there is already knowledge based contamination criteria available.
650		0	_aPlant tissue culture.
650		0	_aAbaca (Plant)
650		0	_aPlant propagation.
650		0	_aComputer algorithms.
655		7	_aAcademic theses.
942			_2ddc _cT&D