Predikciós statisztikai módszerek az ásványvízfogyasztás előrejelzésében

Absztrakt: Magyarorszagon az asvanyvizfogyasztas szokasai az elmult 40 evben gyokeresen megvaltoztak. Az elelmiszerek kozul ez a termekkategoria produkalta a legnagyobb novekedest ezen időszak alatt, ugyanakkor az utobbi evekben a novekedesi utem lelassult. Az asvanyviz kiskereskedelem monitorozasat - fogyasztas mennyisege, kiszereles tipusa, kiszereles merete, markak reszesedese, atlagarak, bolt tipusok reszesedese - az AC Nielsen piackutato vegzi, mig az asvanyvizfogyasztas kedveltseget es gyakorisagat a GfK Piackutato Intezet „Etkezesi Szokasok" cimű kutatasaiban gyűjti. Kutatasunkban jellemezzuk a hazai asvanyvizfogyasztast, valamint statisztikai modszerekkel predikciot vegzunk a fogyasztas varhato alakulasara. Kutatasi kerdeseink a kovetkezők: 1. A predikcios modell becslese melyik evtől nem valtozik szamottevően a tovabbi evek adatainak ismereteben, azaz mikortol műkodik rogzitett parameterekkel is megbizhato előrejelzeskent? 2. A modell alapjan a fogyasztas hogyan alakul az elkovetkező evekben? Az eredmenyeink alapjan 2010 ota a predikcios modell csupan a korabbi evek adataibol is jol becsuli a fogyasztast (RMSE 2011-2016<5 liter/fő/ev). A modell a kozeli evekre igen pontosan becsul (atlagos RMSEel5rcelsoev=3,01, atlagOS RMSEelore masodik ev=3,58, atlagos RMSEelore harmadik ev= 4,42 [liter/fŐ/ev]). A modell alapjan a kovetkező 3 evben atlagosan 0,7 liter/fő/ev lesz a novekedes, utana kovetkező evekben a novekedes uteme lelassul, es a telitődest 127 liter/fő/ev koruli fogyasztasi szinttel eri el. Abstract: The habits of mineral water consumption have radically changed over the last forty years in Hungary. Among the food and beverage goods, this category of product showed the greatest growth during this period, however in the past few years the growth rate had slow down. Monitoring of the mineral water retail - volume of consumption, type of packaging, size of packaging, share of brands, average prices, share of share of market types - is carried out by the AC Nielsen market researcher. The popularity and frequency data of mineral water consumption are collected by the Gfk Market Research Institute for its .Fating Habitats" study. In our research, we characterize the consumption of mineral water, and make a prediction the expected evolution of consumption with statistical methods. Our research questions are the followings: 1. According to the analyzed dataset, from which year does not change considerably the estimate of the prediction model? When does the model works as a reliable forecast with fixed parameters? 2. How consumption evolves over the next few years based on the model? Based on our result, the prediction model gives a very good estimate about consumption since 2010 (RMSE 2011-2016<5 litre/person/year). The model is very accurate for the near years (average RMSE„ext first year=3.01, average RMSEnext second year=3.58, average RMSEnexl mw ycar=4.42 [litre/person/year]). According to this model, the average growth will be 0.7 litre/person/year in the next three years, after that the growth rate had slow down and in the final state the consumption level will be 127 litre/person/year.