Interntionl Journl of Business nd Economics Reserch 2017; 6(4): 67-72 http://www.sciencepublishinggroup.com/j/ijber doi: 10.11648/j.ijber.20170604.15 ISSN: 2328-7543 (Print); ISSN: 2328-756X (Online) Evlution of Grey Forecsting Method of Totl Domestic Coffee Consumption in Indonesi Tien-Chin Wng 1, *, Muhmmd Ghlih 1, 2 1 Deprtement of Interntionl Business, Ntionl Kohsiung University of Applied Sciences, Kohsiung, Tiwn 2 Deprtement of Agriculturl Industril Technology, Stte Polytechnic of Tnh Lut, Pelihri, Indonesi Emil ddress: tcwng@kus.edu.tw (Tien-Chin Wng), ghlih@politl.c.id (M. Ghlih) * Corresponding uthor To cite this rticle: Tien-Chin Wng, Muhmmd Ghlih. Evlution of Grey Forecsting Method of Totl Domestic Coffee Consumption in Indonesi. Interntionl Journl of Business nd Economics Reserch. Vol. 6, No. 4, 2017, pp. 67-72. doi: 10.11648/j.ijber.20170604.15 Received: My 19, 2017; Accepted: June 29, 2017; Published: July 19, 2017 Abstrct: Indonesi is the one of the world s top coffee producing nd exporting countries. Menwhile, in this study, only focus on forecsting the totl of domestic coffee consumption in Indonesi pplied the Grey differentil model which is clled GM (1,1) model of Grey theory to predict the mount of domestic coffee consumption in Indonesi from 1990 to 2017. According to the estimted result, the verge residul error of the Grey forecst model is over 5 percent. The model predicts tht the totl of consumption will increse in ech yer. Bsed on the experimentl results, this proposed method pprently not only improve the forecsting ccurcy of the originl Grey models but lso provide vluble reference for Indonesi coffee frmer nd industries to mke the ction pln for the future. Keywords: Coffee, Consumption, Forecsting, GM (1,1), Grey Theory, Indonesi 1. Introduction Indonesi coffee consumption continues to grow currently, driven in prt by the mrketing strtegies nd brnd wreness [23] efforts. Consumption growth powered by the expnsion of retil coffee shops, including frnchises nd locl smll business. Coffee outlets trget consumers in shopping mlls, business centers nd public fcilities such s irport nd trins sttions. According to dt from the Assocition of Indonesin Coffee Exporter nd Industries [33], Indonesin frmers together with relted ministries pln to expnd Indonesin coffee plnttions, while rejuventing old plnttions through intensifiction progrms. By incresing frm re, Indonesin coffee production in the next ten yers is trgeted to rech between 900 thousnd tons to 1.2 million tons per yer. Due to the incresed globl nd domestic demnd, investment in the country's coffee sector required. In ddition to incresing quntity of coffee bens, qulity is lso predicted to increse due to technologicl innovtions. Nevertheless, Indonesi's coffee production per hectre is still smll compred to the other leding coffee-producing countries. In 2015, Indonesi produced 741 kg of robust seed per hectre nd 808 kg of rbic seed per hectre. In Vietnm, this figure reched 1,500 kg per hectre in Brzil reching 2,000 kg per hectre. In 2012, pproximtely 70 percent of the totl nnul production of Indonesin coffee bens ws exported, especilly to customers in Jpn, South Afric, Western Europe nd the United Sttes. Besides, with totl consumption of 3.6 million bgs in 2012 [32], Indonesi is the second-lrgest consumer in the region, from Jpn, nd the 8 th lrgest in the world. consumption hs been incresing quickly, verging 6.6 percent growth since 2000, nd 5 percent per nnum going bck to 1990. Menwhile, with popultion of nerly 250 million, per cpit consumption is less thn 1 kg per person, nd shows significnt potentil for further growth. Furthermore, it cn ssume tht the mjority of consumption in Indonesi is of
Interntionl Journl of Business nd Economics Reserch 2017; 6(4): 67-72 68 ntionl production, which is 80 percent robust. Furthermore, Indonesi lso imports round 1 million bgs of coffee, predominntly from Vietnm, which further suggests tht most consumption is of robust coffee. domestic consumption s percentge of totl coffee production hs incresed from n verge of 22 percent in the 1990s to round 33 percent over the lst five yers. If the consumption in Indonesi continues to grow t current rtes, the country could rech nerly 6 million bgs by 2020, exceeding the current consumption of Frnce [32]. Thereto, s Indonesi's domestic coffee consumption hs grown, the number of exports hs declined. coffee consumption in Indonesi incresed with compound nnul growth rte (CAGR) of 7.7 percent in 2011 to 2014. Still, t 1.0 kg (dt 2014), coffee per cpit consumption remins low in Indonesi. Wheres, in Asi the totl consumption is resonbly high in Indi, Indonesi nd Philippines, lthough per cpit consumption levels re reltively low for instnce Asi nd the Pcific (estimted) 8.328 such s Indi 1.800, Indonesi 3.333, Philippines 1.080 nd Vietnm 1.583. Further, consumption of instnt coffee mixes nd redy-to-drink beverges is lso growing. This reserch collected 27 yers dt series the totl of rbic nd robust coffee for domestic coffee consumption in Indonesi from 1990 to the 2017 [32]. Moreover, to producing regulr coffee in Indonesi the frmers produces some specilty coffee. The most fmous mong these specilty coffees re Luwk coffee, Torj coffee, Aceh coffee nd Mndiling coffee. The first type of coffee - Luwk coffee - mungking is the most fmous coffee type becuse it is known s the most expensive coffee in the world. This coffee is extrcted from the coffee bens tht hve been through the Asin civet mongoose digestive system (nimls tht resemble cts). Becuse of the specil fermenttion process in the niml's stomch (nd lso becuse the fcts of mongoose cn choose the most juicy coffee fruit) this coffee is believed to hve richer tste. The production process tht requires lot of mnpower nd scrcity in the interntionl mrket cuses the price to be expensive [34]. On the other hnd, longitudinl study by Deng [3, 4, 5, 6] reported tht the first systemtic study of Grey theory in 1982, which hs been recognized nd pplied by mny cdemicins in different subjects such s economic prediction [20, 24, 25, 29, 31], mteril science [11, 12, 14, 15], electricl power [2, 8, 9, 10], trffic [7, 17, 21, 26], technologicl progress [1, 13, 16], engineering [18, 19] nd griculture [22], chosen Grey prediction s n bility forecsting mens becuse of hving reltively low dt requirements, nd GM model constructed from smple of just four pieces of dt. In ddition to tht forecst method is significnt by using the trnsformed Grey rolling modeling mechnism. This rolling modeling mechnism provides mens to gurntee input dt re lwys the most recent vlues from time series dt to forecst the number to get the result. The present pper exmines the Grey forecsting method of totl domestic coffee consumption in Indonesi, during 27 yer since 1990 until 2017 nd forecst to 2018. In the first prt of the pper review of literture regrding the existence of Grey forecsting method on different reserch is presented. It could hve been observed tht the Grey forecsting method used for severl subjects. Further on, the pper presents the methodology nd the dt tht were used, but lso empiricl results tht were obtined for ech observtion period. An expression introducing of rolling modeling dt nd dt of forecst results show their verge residul error different from rolling modeling GM (1,1). This discrepncy my be due to the study presents methodologies for projecting the most correctly predicts of the totl coffee consumption to get the reference for Indonesi to mke the ction pln for the future by nlyzing the precision of the Grey forecsting model. In the end of pper, the conclusion tht resulted from the nlyses re presented, long with the improve the forecsting ccurcy of the originl Grey models, nd provide vluble reference for Indonesi coffee frmer nd industries to mke the ction pln for the future. 2. Reserch Aims nd Previous Studies The fundmentl purposes of the study re mention s follows: Highlight the significnce of Indonesi coffee consumption. Arrnge tht direct reltionship exists between totl coffee consumption nd totl coffee production in Indonesi. Chrcterize lterntive tools in Grey method to mke forecsting useful for decision nd policy mkers need in future prediction of coffee consumption. The proposed reserch is significnt s not only highlights the importnce of Grey forecsting method to Indonesi coffee consumption but provides strtegies tht cn crete knowledge in the totl production of coffee in more cost effective nd efficient mnner to the future. Besides, there re mny studies tht hve ddressed the issue of prediction using Grey forecsting method. This tble below shows severl of longitudinl studies by the other reserchers: Tble 1. The most importnt reserchers using Grey forecsting. Resercher Yer Deng Julong 1982 Xinmin Wng 1999 Chin-Tsi Lin 2003 Sue J. Lin 2007 Chiun-Sin Lin 2011 Rotchn Inthrthirt 2015 Liping Zhng 2017
69 Tien-Chin Wng nd Muhmmd Ghlih: Evlution of Grey Forecsting Method of Totl Domestic Coffee Consumption in Indonesi 3. Methodology A considerble mount of literture hs been published on Grey theory, developed by Deng [3] in 1982, is suitble for short-term forecsting, nd does not rely on sttisticl method. Also, the Grey forecsting method hs been successfully pplied in mny res of reserch including finnce, engineer, griculture nd mngement. Furthermore, in Grey generting system such s Grey reltionl nlysis, Grey forecsting, Grey decision, nd Grey controller re the minly methodology of Grey system theory. However, in this study we focus on the forecsting method is significntly by pplying the trnsformed Grey rolling modeling mechnism. This rolling modeling mechnism provides mens to gurntee input dt re lwys the most recent vlues. In nother mjor study Zheng et l [31], this reserch pplied the generl GM (1,1). In consequence n expression introducing the comprison of rolling modeling dt nd fundmentl dt of forecst results show s Figure 1. The verge residul error different rolling modeling GM (1,1). For instnce, in Method 1: Choose first four continuous dt to predict the 5 th of the output vlue, 2 nd to 5 th consecutive dt to predict the 6 th output vlue nd therefter. Besides, in Method 2: forecst the 6 th of the production vlue by dopting first five consecutive dt, 2 nd to 6 th consecutive dt to forecst the 7 th output vlue nd henceforth. Furthermore, the study presents methodologies for projecting the most ccurtely predicts of the mount of coffee consumption in Indonesi by testing the precision of the Grey forecsting model. Detiled exmintion by Deng [3] proposed the Grey system theory to build Grey model for forecsting. Accumulted Genertion Opertion (AGO): Accumulting obtined systemtic regulrity discrete the time series dt. ( 0 ) (0) (0) (0) x = ( x (1), x (2),..., x ( n)) (1) (1) (0) x is x one-order ccumulted generting opertion (AGO) sequence, tht is, 1 2 n (0) (1) (0) (0) = (2) x ( x ( k), x ( k),..., x ( k)) k= 1 k = 1 k= 1 Inverse-ccumulted generting opertion (IAGO): (0) Gry Derivtives. (1) (1) x ( k) = x ( k) x ( k 1) (1) (1) (1) z = 0.5 x ( k) + 0.5 x ( k 1) (3) Gry Difference Eqution Derivtives. The first order differentil eqution of GM(1,1) model is dx/ dt + x = b, where t denotes the independent vribles in the system, represents the developed coefficient, b is the Grey controlled vrible, moreover nd b denoted the prmeters requiring determintion in the model. When model is constructed, the differentil eqution is 0 (1) x ( k) + z ( k) = b, including k = 2,3,..., n, where, b denoted stndby substntil number, this differentil eqution (0) (1) x ( k) + z ( k) = bis clled s GM (1,1) model. T T T 1 T N =, N =, = ( ) N Y BA B Y B BA A B B B Y Furthermore, ccumulted mtrix nd b re s below expnd equtions: b = = n n n (1) (0) (1) (0) z ( k) x ( k) ( n 1) z ( k) x ( k) k = 2 k = 2 k = 2 n 2 n (1) (1) ( n 1) z ( k) z ( k) k = 2 k= 2 n n n n (1) 2 (0) (1) (1) (0) [ z ( k)] x ( k) z ( k) z ( k) x ( k) k = 2 k = 2 k = 2 k = 2 n n (1) 2 (1) 2 Whitening Eqution: ( n 1) [ z ( k)] [ z ( k)] k = 2 k = 2 (1) (1) b k b (1) b ( k 1) b x ( k) = x (1) e e x (1) e + = + where x (1) (0) b k b x ( k + 1) = x (1) e +, (1) (0) (1) = x (1). Utilize Inverse-ccumulted generting opertion (IAGO) eqution s below: (0) (1) (1) (0) b x ( k + 1) = x ( k + 1) x ( k) = (1 e ) x (1) e 4. Dt Anlysis nd Results 2 k The results of this study indicte tht dt series from Interntionl Coffee Orgniztion [32] used to forecst the totl of coffee consumption in Indonesi. This study produced results tht corroborte the findings of severl previous studies in this field. Wht is noteworthy in Tble 2 is tht the rw dt of coffee consumption in Indonesi from 1990 to the 2017 yer. Focus on 2007 to 2011 hve the similr number, in this cse, we need to modify little number becuse the model cnnot run the sme number. (4) (5) (6)
Interntionl Journl of Business nd Economics Reserch 2017; 6(4): 67-72 70 Tble 2. Totl of domestic coffee consumption in Indonesi (thousnd 60kg). Crop yer Rel Dt Crop yer Rel Dt 1990/91 1,242 2004/05 2,000 1991/92 1,280 2005/06 2,500 1992/93 1,319 2006/07 2,833 1993/94 1,359 2007/08 3,333.00 1994/95 1,400 2008/09 3,333.00 1995/96 1,443 2009/10 3,333.00 1996/97 1,486 2010/11 3,333.00 1997/98 1,532 2011/12 3,667 1998/99 1,578 2012/13 3,900 1999/00 1,626 2013/14 4,167 2000/01 1,676 2014/15 4,333 2001/02 2,000 2015/16 4,500 2002/03 1,779 2016/17 4,600 2003/04 1,833 2017/18 4750* *Forecsting 5percent residul error Source: Interntionl Coffee Orgniztion In Figure 1, we cn see tht the different estimte between the rel dt nd the forecst from Grey model theory. This result explined by the fct tht line chrt in blue color indicte the ctul dt nd the red color indicte the forecst from Grey forecsting. It is pprent in Figure 1 tht the forecst from 5 to 11 yers period remined stble stted tht Grey forecsting method showed the significnt correltion between the rel dt with the Grey forecsting method. The present findings seem to be consistent with the results of previous reserch, showing tht Grey forecsting method cn be used to predict the future of totl domestic coffee consumption in Indonesi. Figure 1. Rolling model for forecsting from 1990 to 2017. An optimum number to forecst the totl of domestic coffee consumption in Indonesi hs been developed with α = 0.4 [7, 8, 16, 28, 31], dt series length m = 4, nd dt series step = 1. Slightly worse prediction results re obtined with α = 0.4, m =4 nd = 12, tht is with the prediction from the sme month of the previous yers. Such dt on GM (1,1) prediction of totl coffee consumption in Indonesi from Tble 2 re given in Tble 3. In ddition, there re the results for four yers period. However, for the purpose of nlysis in Tble 3 we cn see tht the lower error is more thn 5 in 4 yers nd it is lso the error higher round more then 8 in 12 yers. According to the verge residul error tht indicte the dt time series bout totl dt consumption of coffee in Indonesi is suitble to use Grey forecsting method. Tble 3. Averge residul error. Percent Totl Error Percent Totl Error 1-3yer - 8-yer 7.52 4-yer 5.2 9-yer 7.59 5-yer 6.25 10-yer 7.59 6-yer 7.08 11-yer 8.37 7-yer 7.48 12-yer 8.43 5. Prediction Evlution The evlution of GM (1,1) method in this study used to forecst the mount of domestic coffee consumption in Indonesi from 1990 to 2017. The results presented tht the verge ccurcy of the forecsting model exceeds 85 percent. The model thus clerly
71 Tien-Chin Wng nd Muhmmd Ghlih: Evlution of Grey Forecsting Method of Totl Domestic Coffee Consumption in Indonesi hs high prediction vlidity nd is vible gol for forecsting the totl of domestic coffee consumption in Indonesi. Moreover, from the forecst shows tht the totl of coffee consumption in Indonesi will continue to increse, nd it will drive to mke new pln for rod-mpping in the future, nd from the forecst, the result cn count the totl of production coffee from Indonesi to prepre better pln in the future. This lso grees with our erlier observtions, which demonstrted tht result cn explin nd forecst for instnce, in Method 1: choose first four continuous dt to forecst the 5 th of output vlue, 2 nd to 5 th consecutive dt to forecst the 6 th output vlue nd therefter, in Method 2: Predict the 6 th of output vlue by dopting first five consecutive dt, 2 nd to 6 th consecutive dt to forecst the 7 th output vlue nd henceforth. In consequence of result, it is cn be seen from the results shows tht the Grey forecsting model exhibits highest forecst ccurcy nd n verge ccurte rte t the verge residul error of the Grey forecsting model is lmost over 95 percent. The bove sttistics confirm the efficiency of the proposed forecsting model. In such wy, the forecsting method by pplying the Grey rolling model is the most ccurte predictive method to the trend of the totl of domestic coffee consumption in Indonesi. 6. Discussion nd Future Work Grey system modeling from GM (1,1) usully exhibits mximum ccurcy for = 0.5. As cn be seen in Figure 1, the result of forecsting the totl of domestic coffee consumption in Indonesi depends on the originl dt. We cn see the dt in 2007 to 2011 the dt lmost the sme hve some number, nd we should chnge little number to get the result becuse the system cnnot run the sme number. In GM (1,1) used for the demnd of coffee consumption to forecst the number of error nevertheless, the prmeter strts from 5.2 until 8.43 in Tble 2. However, in other cses, the ccurcy of the forecsting depends highly on the dt series used for forecsting. In short, dt series should pproximte with severl chrcteristic signs such s given in Tble 3. In future investigtions, it might be possible to use different Grey forecsting model in totl coffee production in Indonesi bout export nd import. Hence, the Grey prediction method lso will led to nother reserch focus on such the price of coffee in the future with the different trget mrket. Further study smidgen greter focus on export is suggested becuse Indonesi is the one of the world s top coffee producing nd exporting countries. 7. Conclusion The key fetures of the Grey forecsting method re to predict dt series to get future prediction number using previous dt. In this pper, the Grey system modeling bsed on the experimentl results for forecsting the totl of domestic coffee consumption in Indonesi were exmined nd showed the number bout 5.2 percent totl error to predict rnge one until four yers. As result of the test tht ws conducted tht the totl of domestic coffee consumption in Indonesi increses yer by yer ccording to the rel dt nd the Grey forecst method. In conclusion, it cn be sid tht Indonesi government, coffee frmer, nd industries should mke the ction pln for the future to develop coffee sectors in ll spects becuse of the coffee mrket nowdys develops rpidly in over the world. Acknowledgements This reserch supported by the Ministry of Reserch, Technology nd Higher Eduction of the Republic of Indonesi. We pprecite the referees for their kind nd helpful comments. References [1] Chen-Fng Tsi, Dynmic grey pltform for efficient forecsting mngement, Journl of Computer nd System Sciences, 81, 2015, pp. 966 980. [2] Coskun Hmzcebi nd Huseyin Avni Es, Forecsting the nnul electricity consumption of Turkey using n optimized Grey model, Energy, 2014, pp. 165-171. [3] Deng, Julong, Control problems of Grey systems, Systems nd Control letters 5, 1982, pp. 288-294. [4] Deng, Julong, Introduction to Grey system theory, The Journl of Grey System, Vol. 1, No.1, 1989, pp. 1-24. [5] Deng, Julong, The Course on Grey System Theory, Huzhong University of Science & Technology Publish House, Wuhn, Chin, 1990, p. 91. [6] Deng, Julong, The Essentil Methods of Grey Systems, Huzhong University of Science nd Technology Press, Wuhn, Chin, 1992. [7] Hosse, René S, Becker, U, Mnz, H, Grey Systems Theory Time Series Prediction pplied to Rod Trffic Sfety in Germny, IFAC-PpersOnLine 49-3, 2016, pp. 231 236. [8] Huiru Zho nd Sen Guo, An optimized Grey model for nnul power lod forecsting, Energy, 107, 2016, pp. 272-286. [9] Li. C. H, Using improved Grey forecsting models to forecst the output of opto-electronics industry, Expert Syst. Appl. 38, 2011, pp. 13879 13885. [10] Lee, Y. S nd Tong, L. I, Forecsting energy consumption using grey model improved by incorporting genetic progrmming. Energy Convers. Mng. 52, 2011, pp. 147 152. [11] Lee, C, Lin, C. T, Chent, L. H, Accurcy nlysis of the Grey Mrkov forecsting model. J. Stt. Mng. Syst. 7, 3, 2004, pp. 567 580.
Interntionl Journl of Business nd Economics Reserch 2017; 6(4): 67-72 72 [12] Lee, Y. C, Wu, C. H, Tsi, S. B, Grey system theory nd fuzzy time series forecsting for the growth of green electronic mterils. Int. J. Prod. Res, 52 10, 2014, pp. 2931 2945. [13] Li, D et ll, Forecsting short-term electricity consumption using the dptive grey-bsed pproch-n sin cse. Omeg 40 (6), 2012, pp. 767 773. [14] Liu, S., Dng, Y., Fng, Z., Xie, N., Grey System Theory nd Its Appliction. Science Press, Beijing Chin, 2010. [15] Liu, S. nd Lin, Y., Grey informtion: theory nd prcticl pplictions, Springer Science & Business Medi, 2006. [16] Liu S, Forrest J, Yng Y. A brief introduction to Grey systems theory. In: Grey systems: theory nd ppliction, specil issue: selected ppers from the 2011 interntionl conference on Grey systems nd intelligent services (IEEE GSIS), 15-18 September 2011, Nnjing, Chin, vol. 2, 2012. pp. 89-104. [17] Mo, M, Appliction of Grey model GM(1, 1) to vehicle ftlity risk estimtion. Technologicl Forecsting nd Chnge, 73 (5). 2006. [18] Ning Xu, Yoguo Dng, Ynde Gong, Novel Grey prediction model with nonliner optimized time response method for forecsting of electricity consumption in Chin, Energi, 2017, pp. 473-480. [19] Omidvri, M, Presenting model for sfety progrm performnce ssessment using Grey system theory, In Grey System Theory nd Appliction 4(2), S, 2014, pp. 287-298. [20] Jui-Fng Chng, et ll, Forecst the mount of import nd export in Vietnm by pplying Grey Method, ICIC Express Letters, vol. 4, no.5(a), 2010, pp. 1665-1670. [21] Sevl Ene nd Nursel Öztürk, Grey modelling bsed forecsting system for return flow of end-of-life vehicles, Technologicl Forecsting & Socil Chnge, 115, 2017, pp. 155 166. [22] Shng-Ling Ou, Forecsting griculturl output with n improved grey forecsting model bsed on the genetic lgorithm, Computers nd Electronics in Agriculture 85, 2012, pp. 33 39. [23] Tien-Chin Wng, Muhmmd Ghlih, Glen Andrew Porter, Mrketing Public Reltions Strtegies to Develop Brnd Awreness of Coffee Products, Science Journl of Business nd Mngement. Vol. 5, No. 3, 2017, pp. 116-121. [24] Tien-Chin Wng, Su-Hui Kuo, Hui-Chen Chen, Forecsting the Exchnge Rte between ASEAN Currencies nd USD, Industril Engineering nd Engineering Mngement (IEEM), IEEE, 2011. [25] Tien-Chin Wng, et l, Forecst the foreign exchnge rte between Rupih nd US Dollr by pplying Grey Method, Interntionl Conference on Dt Engineering nd Internet Technology, 2011, pp. 550-553. [26] Wng, L, Forecsting of Trffic Accident in Shnxi Province bsed on grey system theory, In 2nd Interntionl Conference on Remote Sensing, Environment nd Trnsporttion Engineering, Nnjing, Chin, 2012. [27] Ximin Wng, et l, Grey predicting theory nd ppliction of energy consumption of building het - moisture system, Building nd Environment, 1999, pp. 417-420. [28] Xiuli Liu, Blnc Moreno, An Slome Grcí, A Grey neurl network nd input-output combined forecsting model. Primry energy consumption forecsts in Spnish economic sectors, Energy, 115, 2016, pp. 1042-1054. [29] Ynhui Chen, et l, Multi-step-hed Crude Oil Price Forecsting bsed on Grey Wve Forecsting Method, Procedi Computer Science, 2016, pp. 1050-1056. [30] Ynhui Chen, Kijin He, Chun Zhng, A novel Grey wve forecsting method for predicting metl prices, Resources Policy, 49, 2016, pp. 323 331. [31] Zheng-Xin Wn nd De-Jun Ye, Forecsting Chinese crbon emissions from fossil energy consumption using non-liner Grey multivrible models, Journl of Clener Production, 2017, pp. 600-612. [32] Informtion on http://www.ico.org/. [33] Informtion on http://www.eki-ice.org/. [34] Informtion on https://www.indonesi-investments.com/. [35] Informtion on http://www.remrkbleindonesincoffee.com/.