=Paper=
{{Paper
|id=Vol-1152/paper43
|storemode=property
|title=Application of Data Mining Techniques to Olea Europaea var. Media Oblonga Production from Thassos Island
|pdfUrl=https://ceur-ws.org/Vol-1152/paper43.pdf
|volume=Vol-1152
|dblpUrl=https://dblp.org/rec/conf/haicta/TheodosiouVHN11
}}
==Application of Data Mining Techniques to Olea Europaea var. Media Oblonga Production from Thassos Island==
https://ceur-ws.org/Vol-1152/paper43.pdf
Application of Data Mining Techniques to Olea
europaea var. media oblonga production from Thassos
Island
Theodosios Theodosiou1, Stavros Valsamidis1,Georgios Hatziliadis1, Michael
Nikolaidis1
1
Department of Accountancy, Technology Educational Institute of Thessaloniki, Greece
e-mail: theodosios.theodosiou@gmail.com, {svalsam,mnikol}@teikav.edu.gr, egs-
kav@otenet.gr
Abstract. A huge amount of data is produced in our days in the agriculture
sector. Due to the huge amount of this datasets it is necessary to use data
mining techniques in order to comprehend the data and extract useful
information. In our work we apply three different data mining techniques to
data about Olea europaea var. media oblonga from the island of Thassos, at
the northern part of Greece. The data were from 1063 farmers from three
different municipalities of Thassos, namely Kallirachi, Limenaria and Prinos
and concerned 2010. They were analysed using the classification algorithm
OneR, the clustering algorithm k-means and the association rule mining
algorithm, Apriori from the WEKA data mining package. The results indicate
that organic cultivation could improve the production of olives and olive oil.
Keywords: Olea europaea, Classification, Clustering, Association Rule
Mining.
1 Introduction
Nowadays huge volumes of agricultural data are accumulated from a variety of
sources on almost daily bases. The size of the datasets makes obsolete manual
analysis and extraction of useful information. Consequently data mining techniques
are becoming a necessity in order to process and analyse these datasets. Data Mining
is an iterative process of creating predictive and descriptive models, by uncovering
previously unknown trends and patterns in vast amounts of data, in order to extract
useful information and support decision making (Mucherino et al., 2009).
The application of data mining techniques into research areas such as the
agricultural field is quite an emerging area of research. The techniques applied to
agriculture are not specifically designed for this field. On the contrary they are quite
general and could be easily applied to any type of data. (Mucherino et al., 2009).
Abdullah et al. (2004) applied data mining in order to discover rules for the use of
________________________________
Copyright ©by the paper’s authors. Copying permitted only for private and academic purposes.
In: M. Salampasis, A. Matopoulos (eds.): Proceedings of the International Conference on Information
and Communication Technologies
for Sustainable Agri-production and Environment (HAICTA 2011), Skiathos, 8-11 September, 2011.
487
pesticides in agriculture. Data mining methods are divided into three major
categories. The first category involves the classification methods, whereas the second
the clustering ones and the third the association rule mining methods.
Classification methods use a training dataset in order to estimate some parameters
of a mathematical model that could in theory optimally assign each case from a new
dataset into a specific class. In other words, the training set is used to train the
classification technique how to perform its classification.
Clustering refers to methods where a training set is not available. Thus, there is
no previous knowledge about the data to assign them to specific groups. In this case,
clustering techniques can be used to split a set of unknown cases into clusters.
Association rule mining discovers relationships, sometimes hidden, among
attributes (variables) in a dataset.
In our research we apply methods from these three data mining major categories
to cultivation and production data of Olea europea var. media oblonga from the
island of Thassos. Olive oil is a natural fruit juice with excellent nutritional
characteristics. It is a typical lipid source food of the Mediterranean diet and its
consumption has been associated with a low incidence of cardiovascular diseases,
neurological disorders, breast and colon cancers, as well as with hipolipidemic and
antioxidant properties. An increase of interest in olive oil as a healthy food has been
observed lately in areas other than the Mediterranean countries mainly because of its
fatty acid composition and content of other functional food components, such as
polyphenols (Vekiari et al., 2010).
Olive cultivation is exceptionally spread in the island of Thassos, in Greece. The
variety of olive cultivated all over Thassos is Throumbolia (Olea europaea var. media
oblonga). The Throumbolia variety grows at altitudes of up to 700m and its fruits are
medium-size. Its main characteristic is that under special conditions of temperature
and moisture, the bitter taste, which is evident in the olives while they are still on the
tree, disappears due to the hydrolysis of oleuropein by the action of the fungus
Phoma oleae. It is important that olives are harvested as soon as they fall from the
tree and pressed immediately to produce sweet oil, rich in aromatic substances,
because of the olive trees height. Otherwise, a deterioration of olive quality usually
occurs resulting in unpleasant tasting oil (Vekiari et al., 2010).
It is well known that the differences in olive oil quantity from various regions are
attributed to olive variety, environmental factors, harvesting methods, time of
harvest, and extraction techniques. Furthermore, there is little information on the
specific factors that influence the variety of Throumbolia on the island of Thassos.
Since, olive variety, harvesting methods, time of harvest and extraction techniques
for public extraction factories are almost identical for the farmers in our study, we
would like to investigate the difference in the environmental factors if and how they
affect olive oil quantity. The trees are cultivated in three different areas of Thassos,
namely Kallirachi, Limenaria and Prinos. Limenaria is considered to have a drier
climate compared to the other two.
Furthermore, we also investigate the effects of organic cultivation or biological
cultivation, as it is usually called in Greece, on the quantity of olive oil. Organic
cultivation is the form of agriculture that relies on techniques such as crop rotation,
green manure, compost and biological pest control to maintain soil productivity and
control pests on a farm. Organic cultivation excludes or strictly limits the use of
488
artificial fertilizers, pesticides (which include herbicides, insecticides and
fungicides), plant growth regulators such as hormones, livestock antibiotics, food
additives, and genetically modified organisms1. The latest years there are an
increasing number of farmers that are changing their type of cultivation in the island
of Thassos from conventional ones to organic.
Section 2 describes the dataset and the methodology used in our research.
Sections 3 presents the results from the data mining methods and Section 4 discusses
the results and refers to the main conclusions of our research.
2. Data and Methodology
In this section the dataset we used in our methodology is described in detail.
Also, the data mining methods applied to the olive oil data are explained and
analysed.
2.1 The Dataset
The dataset was collected from the Enosi Agrotikon Sinetairismon (EAS) of
Kavala2. The data were collected during 2010 and involve 1063 farmers and 3
different municipalities from the island of Thassos, namely Kallirachi, Limenaria and
Prinos. Prinos and Kallirachi are at the northern part of Thassos whereas Limenaria
at the southern part. The data are originally in ASCII form and are obtained from the
EAS data repository. Each farmer is described by 8 variables in the repository of
EAS. The first two variables are the AFM (tax id) and the last name of the farmer
and they are omitted from our analysis, since they do not influence the production of
olive oil and are private data. The remaining six variables are used in the analysis
described in the methodology section and are called Area, Trees, Bio, Mun, Olives
and Oil. Table 1 describes each variable in detail.
Table 1. The variables used in our analysis
Variable Name Description Type
Area The total area of land owned by the farmer in m2 Numeric
Trees The total number of trees owned by a farmer Numeric
Bio The type of cultivation (0: Typical, 1: Biological/Organic) Numeric
Mun The municipality in which the farmer has his trees Nominal
Olives The number of olives each farmer has Numeric
Oil The total liters of olive oil produced per farmer Numeric
1
Directorate General for Agriculture and Rural Development of the European Commission,
http://ec.europa.eu/agriculture/organic/organic-farming/what-organic_en
2
EAS Kavalas - http://www.easkavalas.gr/
489
2.2 Data mining methods
The WEKA (Waikato Environment for Knowledge Analysis) (Witten & Frank,
2005) computer package was used in order to apply classification, clustering and
association rule mining methods to the dataset. WEKA is open source software that
provides a collection of machine learning and data mining algorithms. Fig. 1 shows
the basic Graphical User Interface (GUI) of WEKA. One of the main objectives of
WEKA is to mine information from existing agricultural datasets (Cunningham and
Holmes, 1999) and the main reason for choosing it for analyzing our data.
Fig. 1. WEKA environment
There are various classification methods implemented in WEKA, like Naïve
Bayes, ZeroR, OneR, etc. In the classification step, the algorithm OneR (Witten &
Frank, 2000) was applied to our data. It uses the minimum-error attribute for
prediction, discretizing numeric attributes (Holte, 1993). The main advantage is that
it produces very simple rules for classification and can be considered the baseline for
classification performance. It was found to perform as well as more sophisticated
algorithms when applied to many of the standard machine learning test datasets
(Holte, 1993). OneR can parsimoniously discover and represent simple relationships
between the real data (Cunningham and Holmes, 1999). In our case the variable
“mun” which shows the municipality the trees are cultivated in is used as a class.
The clustering step uses the k-means algorithm (MacQueen, 1967; Kaufmann &
Rousseeuw, 1990), called SimpleKMeans in WEKA. K-means is an efficient
partitioning algorithm that decomposes the data set into a set of k disjoint clusters. It
is a repetitive algorithm in which the items are moved among the various clusters
until they reach the desired set of clusters. With this algorithm a great degree of
490
similarity for the items of the same cluster and a large difference of items, which
belong to different clusters, are achieved. The Euclidean distance is used to compute
the differences between the olive trees cultivations. The variable “bio” is used in
order to assess the accuracy of the clustering and investigate its impact on olive tree
cultivation.
Association rule mining is one of the most well studied data mining tasks. It
discovers relationships among attributes (variables) in datasets, producing if-then
statements concerning attribute-values (Agarwal, Imielinski, & Swami, 1993). An
association rule X ⇒ Y expresses a close correlation among items in a dataset, in
which transactions in the dataset where X occurs, there is a high probability of
having Y as well. In an association rule X and Y are called respectively the
antecedent and consequent of the rule. The strength of such a rule is measured by
values of its support and confidence. The confidence of the rule is the percentage of
transactions with antecedent X in the dataset that also contain the consequent Y. The
support of the rule is the percentage of transactions in the dataset that contain both
the antecedent and the consequent Y in all transactions in the dataset.
The WEKA system has several association rule-discovering algorithms available.
The Apriori algorithm (Agarwal et al., 1996) is used for finding association rules
over the discretized LMS data table in Appendix 1. Apriori (Agrawal, 1994) is the
best-known algorithm to mine association rules. It uses a breadth-first search strategy
to counting the support of item sets and uses a candidate generation function, which
exploits the downward closure property of support. Iteratively reduces the minimum
support until it finds the required number of rules with the given minimum
confidence.
There are different techniques of categorization for association rule mining. Most
of the subjective approaches involve user participation in order to express, in
accordance with his/her previous knowledge, which rules are of interest. One
technique is based on unexpectedness and actionability (Liu et al, 1996; Liu et al,
2000). Unexpectedness expresses which rules are interesting if they are unknown to
the user or contradict the user’s knowledge. Actionability expresses that rules are
interesting if users can do something with them to their advantage. The number of
rules can be decreased to unexpected and actionable rules only (García et al., 2008).
Another technique proposes the division of the discovered rules into three categories
(Minaei-Bidgoli et al., 2004). (1) Expected and previously known: This type of rule
confirms user beliefs, and can be used to validate our approach. Though perhaps
already known, many of these rules are still useful for the user as a form of empirical
verification of expectations. For agriculture, this approach provides opportunity for
rigorous justification of many long held beliefs. (2) Unexpected: This type of rule
contradicts user beliefs. This group of unanticipated correlations can supply
interesting rules, yet their interestingness and possible actionability still requires
further investigation. (3) Unknown: This type of rule does not clearly belong to any
category, and should be categorized by domain specific experts.
3. Results
The first step before applying the data mining methods described in the previous
section is the pre-processing of the data in order to prepare them for data analysis.
491
3.1 Pre-processing
The three municipalities in the variable “mun”, namely Kallirachi, Limenaria and
Prinos were replaced with the values K, L and P respectively. Furthermore, certain
filters were applied to the data, such as the filter NumericalToNominal in order to
convert numeric variables and their values to nominal. For example, number 0 and 1
in variable Bio are converted to nominal, where 0 signifies conventional cultivation
and 1 organic.
Fig. 2 depicts all the variables used in our analysis. The different scales of gray
correspond to the three different municipalities. Light gray corresponds to Kallirachi,
medium gray to Prinos and dark gray (black) to Limenaria. One can see that 555
farmers where from the area of Limenaria, 297 from Prinos and 211 from Kallirachi.
Also it is evident that for variables area, trees, olives and oil the distribution of their
values is right-skewed. It is noteworthy that only 197 farmers use organic cultivation
and that most of them are in the area of Limenaria.
Fig. 2. Visualization of the attributes with class variable “Mun” (municipality)
3.2 Classification
In the classification step, the algorithm OneR is applied. The attribute “Mun”
(municipality) is used as a class. Fig. 3 presents the overall accuracy of the model
computed from the training dataset and is equal to 77.89%. The worst performance
based on the F-measure that combines precision and recall is for the municipality of
492
Kallirachi and equals 65.2%, whereas the best performance is for the area of
Limenaria and equals 84.2%.
The results indicate that the best attribute, which describes the classification, is
variable Olives that holds the number of olives for each farmer. This means that
variable Olives is more closely related to variable Mun than the other variables and
therefore in some Mun (municipalities) there is higher olives production (Kallirachi,
Prinos) than in other municipalities (Limenaria). A possible explanation for these
results is that the area of Limenaria has a dryer climate compared to Prinos and
Kallirachi and thus the olive trees have lower production.
Fig. 3. Classification results using variable “Mun” (municipality) as class
3.3 Clustering
The clustering step was performed using the k-means algorithm (SimpleKmeans
in the context of WEKA). The number of clusters is set to 2, since the variable “bio”
was used to compute the accuracy of the clustering and inspect the impact of the type
of the cultivation to olive oil production.
Fig. 4 shows the results of the clustering. The incorrectly clustered instances
(farmers) are 26.06% based on variable “bio”. It is also evident from the cluster
centroids that organic cultivation, represented as cluster 0 in the results, contains
trees that produce higher quantities of olive oil compared to cluster 1 (conventional
cultivation). This preliminary result indicates the potential of organic cultivation for
improving olive oil production.
493
Fig. 4. Clustering results. Variable “bio” is used for assessing the clustering.
3.4 Association rule mining
The Apriori algorithm (Agarwal et al., 1996) was used for finding association
rules for our dataset. The algorithm was executed using a minimum support of 0.1
and a minimum confidence of 0.9, as parameters. WEKA produced a list of 10 rules
(Fig. 5) with the support of the antecedent and the consequent (total number of items)
at 0.1 minimum, and the confidence of the rule at 0.9 minimum (percentage of items
in a 0 to 1 scale).
The application of the Apriori algorithm for association provided some
interesting outcomes for the production of olive trees. Figure 5 shows the association
rules that can be discovered. There are of course some uninteresting rules, like rules
2 and 4. They present relatively known information since it is an expected or
conforming relationship between variables Olives and Oil. There are also a couple of
symmetrical rules, since the antecedent element and the consequent element are
interchanged. There is also a similar triad of rules, rules with the same element in
antecedent and consequent but interchanged, such as rules 1, 3 and 7. It is rather
surprising the outcome of rules 5, 6, 7, 8 and 9. Conventional (non-organic) is the
preferable cultivation for farmers with small area (and trees), and consequently small
production of olives and olive oil. The last remark might be useful for the motivation
that might be given to farmers with limited number of trees (or area) to follow the
organic cultivation.
494
Fig. 5. The Apriori algorithm results based on the confidence metric
Summarizing the results from the classification, the clustering and the association
rule mining methods we can conclude that:
(i) The attribute which best describes the classification is the variable Olives that
holds the number of olives for each farmer. The attribute “Mun” (municipality) is
used as a class.
(ii) Using “bio” as class attribute in clustering, namely if cultivation is
conventional or organic, the results show that trees which belong to the second
cluster produce higher quantities of olive and oil compared to the ones with
conventional cultivation.
(iii) In association rule mining, although there are some trivial rules, namely
expected and previously known, like rules 2 and 4 that show that the productions of
oil and olives are mutually dependent, there are also rules like 1, 3, 8, 9 and 10 which
offer a lot of actionability. Organic cultivation has great impact to the olives and oil
production. Rules 5 and 6 show that farmers with limited number of trees and/or area
prefer the organic cultivation.
Overall, the production of olives and oil are based on two different factors. The
first one is environmental and depends on the part of the island that the olives trees
are cultivated. The second factor is the type of cultivation, traditional or organic.
4. Conclusions
The application of three different data mining techniques in agricultural data is
presented in this paper. The dataset involves Olea europea var. media oblonga that is
cultivated in the island of Thassos. It is quite a commercially successful Olea species
that the latest years has been extensively monitored by the EAS of Kavala. The
495
parameters which were investigated were the total area of land owned by the farmer
in m2, the total number of trees owned by a farmer, the type of cultivation (0:
Typical, 1: Biological/Organic), the municipality in which the farmer has his/her
trees, the number of olives each farmer has and the total liters of olive oil produced
per farmer.
The results show some interesting outcomes. First of all, the classification method
indicate that the environmental differences between the northern and southern part of
the island can influence the production of olives and oil, concerning the number of
olives and the quantities of olive oil they can produce. This is mainly due to the fact
that the southern part of the island is considered to be dryer than the northern part
and thus the amount of waterfall influences olive trees. Furthermore, the clustering
results show that organic cultivation could improve olive and oil production.
Although, organic cultivation is usually more cumbersome than conventional one, it
could worth the extra effort if olive oil production is increased. The last data mining
method, the association rule mining, suggests that farmers with limited number of
trees (or area) could benefit more by following the organic cultivation. Organic
cultivation could help them produce more olives and olive oil. In other words organic
cultivation seems to help improve production in the regions presented in the study at
the island of Thassos.
The study has an inherent advantage as far as the data analysis is concerned. The
tree Olea offers two variables as input, the total number of trees and the total area of
land and two variables as output, the number of olives and the total volume of olive
oil.
Further investigation is still required since the results are based on only three
municipalities. In the future we plan to extend the study to other municipalities and
areas of the island Thassos. Furthermore, we plan to overcome the limitation of
manually mining such datasets, by developing a plug-in tool for WEKA to automate
the whole procedure. It should be mentioned that even if the scope of the method is
on olive trees production, it could be easily adopted to the production of other
products of agriculture. A comparative analysis for the same products among
different parts of Greece could also be useful, i.e. for Olea among Thassos and
Lesvos islands and Chalkidiki peninsula.
References
1. Abdullah, A., Brobst, S., Pervaiz, I., Umer, M., Nisar, A. (2004). Learning
dynamics of pesticide abuse through data mining, Proceeding ACSW Frontiers
'04 Proceedings of the second workshop on Australasian information security,
Data Mining and Web Intelligence, and Software Internationalisation - Volume
32.
2. Darcy Miller, Jaki McCarthy, Audra Zakzeski, (2009) A Fresh Approach to
Agricultural Statistics: Data Mining and Remote Sensing, Section on Government
Statistics – JSM 2009
496
3. Agrawal, R. & Srikant, R. (1994). Fast algorithms for mining association rules.
Proceedings of 20th International Conference on Very Large Data Bases (pp.
487-499).
4. Holte, R.C. (1993). Very simple classification rules perform well on most
commonly used datasets. Machine Learning. 11,63-91.
5. Kaufmann, L., Rousseeuw, P.J. (1990). Finding Groups in Data: An Introduction
to Cluster Analysis, New York, John Wiley & Sons.
6. Liu, B. & Hsu, W. (1996). Post-Analysis of Learned Rules. Proceedings of
National Conference on Artificial Intelligence. Portland, Oregon, USA, (pp. 828–
834).
7. Liu, B., Hsu, W., Chen, S. & Ma, Y. (2000). Analyzing the Subjective
Interestingness of Association Rules. IEEE Intelligent Systems, 15(5), 47–55.
8. MacQueen, J. (1967). Some methods for classification and analysis of
multivariate observations. In Proceedings of the fifth berkeley symposium on
mathematical statistics and probability, ( pp. 281–297). California, USA.
9. Minaei-Bidgoli, B., Tan, P-N. & Punch, W.F. (2004). Mining Interesting Contrast
Rules for a Web-based Educational System. Proceedings of Int. Conf. on
Machine Learning Applications, Louisville, USA 2004 (pp. 320- 327).
10. Mucherino, A., Papajorgji, P., Pardalos, P.M. (2009). A survey of data mining
techniques applied to agriculture. Oper Res Int J, DOI 10.1007/s12351-009-0054-
6.
11. Cunningham, S. J., and Holmes, G. (1999). Developing innovative applications in
agriculture using data mining. In the Proceedings of the Southeast Asia Regional
Computer Confederation Conference.
12. Vekiari, S.A., Oreopoulou, V., Kourkoutas, Y., Kamoun, N., Sallem, M.,
Psimouli, V., Arapoglou, D. (2010). Characterization and seasonal variation of
the quality of virgin olive oil of the Throumbolia and Koroneiki varieties from
Southern Greece. Grasas Y Aceites, 61 (3), pp. 221-231.
13. Witten, I. & Frank, E., (2005). Data Mining Practical Machine Learning Tools
and Techniques, San Francisco: Morgan Kaufmann.
497