Agricultural Research Management Information System

Home
Research Summary
Search
About Us
- ARMIS
- Brochure
Contact Us
Report
User Request
Data Input
Help
- Operation Manual
  - PDF
  - Video
- Program Area & Commodity
We have reached 37600 number of research entries at this moment.

- Logout

Research Detail

Home
Research
Detail

Simple Method for Classification of Mango Varieties on the Basis of Their Physicochemical Properties by Chemometric Techniques

M. N. Uddin*
BCSIR Laboratories Dhaka, Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka-1205, Bangladesh

R. Ara
Department of Food Engineering and Tea Technology, Shahjalal University of Science and Technology, Sylhet-3114, Bangladesh

M. Motalab
Institute of Food Science and Technology (IFST), Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka-1205, Bangladesh

A. N. M Fakhruddin
Department of Environment Science, Jahangirnagar University, Dhaka-1342, Bangladesh

B. K. Saha
Institute of Food Science and Technology (IFST), Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka-1205, Bangladesh

Abstract/ Executive Summary:

Nutritional parameters vary significantly among different varieties of mango. It is therefore necessary to have simple method to classify mango pulps according to their nutritional parameters by effective and economic technique. The present study was carried out to develop a method to classify mango varieties by using two chemometric techniques namely, Artificial Neural Network (ANN) and Linear Discrimination Analysis (LDA). At first, 9 physico-chemical parameters have been chosen from 18 of them by applying Factor Analysis (FA), as quantification of each parameter involves time and cost. Nine varieties of mango available in Bangladesh were studied her. LDA can classify the mango pulps on the basis of their nutritional properties 100 percent accurately, and this rate for ANN is 96.3 percent. Therefore, a method is being proposed where mango can be classified with their 9 physico-chemical parameters into their right varieties by LDA. The proposed chemometric method could be used for regular classification of mango pulps at laboratory and mango product manufacturing industries to improve the quality of mango products.

Keywords: Mango varieties; Artificial neural network; Linear discriminant analysis

Location: BCSIR Laboratories Dhaka, Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka-1205, Bangladesh

Start Date:

End Date:

Program Area: Postharvest and Agro-processing

Commodity: Mango

Objectives:

The objective of the study was to assess the classification performance of ANN and LDA and to compare their efficiencies. Therefore, the study was aimed to develop chemometric method for rapid classification of mango to their true varieties on the basis of their physico-chemical properties by using better performing model.

Methodology:

Sample collection Nine popular varieties of mango available in Bangladesh such as, Gopalbhog, Langra, Fazlee, Tosha, Khersapat, Aamrupali, Mollika, Kohitor and Himsagor were collected from local markets of Dhaka City during pick days of their seasons. Nine samples from each variety were taken and thus 81 samples were used finally in the study. Measuring parameters The collected mangoes were washed with de-ionized water, pilled and pulps were taken. Nutritional properties of these mango pulps were determined at Fruit Technology Research Laboratory of Institute of Food Science and Technology (IFST) under Bangladesh Council of Scientific and Industrial Research (BCSIR), Dhaka by using suitable instruments for measuring different parameters. Finally, 18 physico-chemical properties of the mango pulp samples were measured, and they are Edible portion, Moisture content, pH, Titratable acidity, Total Soluble Solids (TSS), Total sugar, Reducing sugar, Ash, Vitamin C, Total Protein, Total Fat, Crude fibre, Total energy, Total carbohydrate, Sodium (Na), Potassium (K), Calcium (Ca) and Magnesium (Mg). Dimension reduction by Factor Analysis (FA) All parameters are not equally important for classification of mango to their varieties. So dimension of the data could be reduced by selecting comparatively more important parameters of mango pulps. In order to reduce the dimensionality of a dataset, FA is very popular and powerful technique. In most of the scientific research variable are large in number and are correlated. It is very often necessary to reduce the dimensionality of the dataset retaining the variability presented in it as much as possible and with a minimum loss of information (Hair et al,1998). This reduction is achieved by transforming the dataset into a new set of variables-factors, which are orthogonal (non-correlated) and are arranged in decreasing order of importance. FA can be expressed as Fi=a1xij+a2x2j+...+amxm Where Fi= factor, a= loading, x= measured value of variable, i= factor number,j = sample number, m= total number of variables Before dimension reduction by FA, it is necessary to test sample adequacy by Kaiser-Mayer-Olkin (KMO) test and Sphericity test for assess eligibility of using FA. Then factors are extracted. Next, rotation of the extracted factors are performed to choose comparatively more important variable and thereby dimensionality of dataset is reduced. Finally, these selected variables are used for further application of chemometric techniques. Artificial neural network (ANN) ANN is a mathematical representation inspired by the human brain, and has an ability to adapt on the basis of the inflow of new information. Mathematically, ANN is a non-linear optimization tool. The ANN is especially suitable for classification and is widely used in practice. The network consists of one input layer, one or more hidden layers and one output layer, each consisting of several neurons. Each neuron processes its inputs and generates one output value that is transmitted to the neurons in the subsequent layer. Initially, a weighted sum of inputs is calculated at each neuron: the out put value of each neuron in the proceeding network layer times the respective weight of the connection with that neuron.

Source of Information: Bangladesh J. Sci. Ind. Res.51(4), 253-260, 201

Article Link (Online Journal):

Funding Source:

Budget:

Total Budget (Tk.) :

Achievement/Findings:

As nutritional values and tastes of mango vary greatly according to their varieties, classifying varieties of mango and its products is a great concern of consumers and producers. Mango could be classified on the basis of nutritional and physic-chemical parameters to their original varieties. Cost, time and efforts are involved with getting measured values of each parameter. So, selection of appropriate physicochemical parameters is one of the most important tasks before classification of mango varieties. The study used Factor Analysis to reduce the dimension of the data. This method efficiently reduced 18 physiochemical parameters to its half. Then Artificial Neural Network, a nonlinear prediction method was used to classify the varieties of mango with 9 selected parameters. ANN could classify mangoes with 3.7 percent misclassification error both in training and test data. Next, another method, Linear Discriminant Analysis, was used for classification of mango, and this method classifies each specimen mango to their correct varieties. Therefore, 96.3 percent of varieties are correctly classified by ANN whereas classification rate by LDA is 100 percent in this regard. Finally, a new analytical method is being proposed where mango pulp could be classified to their respective varieties simply by using measurement values of 9 parameters (total energy, crude fibre, total fat, TSS, total protein, Ash, total carbohydrate and moisture content) and LDA. This method could contribute to improve the nutritional quality of processed commercial mango products like juice, jam, jelly, pickles etc through selection of appropriate varieties of mango.

Knowledge Management: Journal