In this research, we will try to provide effective locational benefits for selling mangoes in Dhaka city. Jahurul et al. have mentioned that nearly 25-30% of mangoes every year go to waste due to weaknesses in postharvest management (Jahurul et al., 2015). Moreover, Yeasmin et al. have notified through a report produced by Hortex Foundation that nearly 24% of mangoes priced at BDT 3600 Crore go to waste each year due to underqualified production management (Yeasmin et al., 2021). Sampa et al mentioned that a majority of mango fruit goes to waste during their transportation to Dhaka which remains the main selling point during the production season of the country (Sampa et al., 2019). Investigations performed by Khandokar et al have shown that due to the high population concentration in Dhaka and with more purchase capability mango farmers expect to sell their products in Dhaka only (Khandokar et al., 2017). Sarker et al. have shown that there has been a slow but gradual shift from crop cultivation to Mango production due to increased profit expectations in the arable lands (Sarker et al., 2014). This certainly proves that Mango could be a beneficial product for the farmers and also for the national economy. Mitra has shown that despite Mango being a tropical fruit its demands are ever increasing as a direct consumable food and also a value-added intermediary food item throughout the entire world gradually (Mitra, 2015). India is the top raw mango exporter, and Thailand is the biggest exporter of value-added mango products, it is observed that in the last 10 years, there has been a gradual increase in sales for both (Mitra, 2015). Hence our suggested research is concentrated on bringing to light the effective evidence of converting Mango as an important product that impacts the national economy of Bangladesh. 

Research Objective 

Our research involves identifying collaborative shreds of evidence regarding Mango production, distribution, post-production management, value addition, and macroeconomic events to efficiently reduce Mango wastage and increase raw mango export both locally and internationally. 

1. Using Locational parameters and mango distribution and selling processes, identify excess mango production amount and using supply chain optimization principles distribute a portion of the excess mangoes for a cheaper price for the overall mango wastage reduction 
2. A portion of excess mangoes should be initialized into value addition in terms of value-added finished goods. This process will effectively increase the longevity of now-wasted mangoes as additional market-demanding products. 

Research Contribution

Currently, research suggests that nearly 25-30% of mangoes go to waste. We would initialize the following contributions in this research to minimize this waste with increasing profit for producers and positive economic benefit for Bangladesh.

1. Shift in demand is evaluated in this research by increasing the product availability and reducing mango prices in the market with initiated pilot projects. 
2. Further evaluations are performed for excess mangoes for conversion into value-added products which are judged using supplementary projects. 
3. Economic benefits for Bangladesh are evaluated as well in terms of FDI (Foreign Direct Investment) as exports increase using Mango. 

Matin et al. have shown that currently there are 5 different exchanging hands of mangoes starting from producers to customers (Matin et al., 2008). Its observed by Matin et al that when so many hands are exchanging such a demanding product, the final price of the product becomes extremely high. Biswas et al. have shown that by using fewer hands, the final price of the product becomes less and the profit-earning opportunity for each hand is reduced. Similarly, Billal et al. have shown that an optimal supply chain network will have less exchange of products for price throughout the supply chain network (Billal et al., 2015). 

Similarly, Azim showed that PRAN by making a purchase from local producers and distributing it at the very end to big retailers reduced exchanging hands thus increasing per unit profit through supply chain optimization (Azim, 2016).  According to BBS data sources 2020, nearly 25 lac tones of mangoes are produced in Bangladesh during its cultivation season where 25% of it goes to waste. Sultana et al suggested that nearly 18% of products face commercial failure due to inefficient product distribution plans, in one of the most popular production districts of Bangladesh named Chapainawabgonj (Sultana et al., 2018). Also, in southern districts like Khagrachori, Bandorban, and SaBDThira, Mango is produced in huge amounts, and is found to make around 35% inefficient selling due to a lack of marketing experience, knowledge of the market, efficient and supportive supply chain network and proper distribution planning (Hossain et al., 2020). As khandoker et al. have priorly mentioned that farmers are shifting their production preference from crops to Mango, then it certainly shows that Mango has become a more profitable product. Ahmed et al. have found that in current days the increase in supply chain instability and increasing prices of all the main constituents of the supply chain including oil price, import dependency, an increase of wage, increase in maintenance and durability costs with hiked storage price plays a crucial role in Mango price increase (Ahmed et al., 2019). Roy & Rahman have observed that in Bangladesh extremely poor and poor mango producers find lower or very limited margins of profit due to fewer products being sold from their end (Roy & Rahman, 2022). Although its observed that extremely poor and poor producers account for nearly 46% of the overall market due to shifts in five different hands, they ultimately gain less profit (Roy & Rahman, 2022). Roy & Rahman further mentioned that by improving strategies for harvesting, grading, packaging, trans-portation, marketing, and overall supply chain distribution mangoes could be sold more efficiently. 

Ahmed has shown that to make efficient production various initiatives like the usage of planting techniques and funding process should be adopted (Ahmed. KU, 1985). Karim showed that despite various adopted techniques that impact the management quality of the mango industry, marketing and distribution efficiency is still lacking (Karim. A, 1985). To classify the relevant issues for efficient mango distribution various factors like supply increment, price decrease, demographics of selling point, ease of commute, etc. issues are required to be integrated (Hudnukar et al., 2014). For such situations, its better to apply an unbiased algorithm proposed by machine learning methods known as K Means Clustering (Hartigan & Wong, 1979). Moreover, K Means clustering successfully segregated qualified stocks from underqualified stocks using unbiased and unsupervised classification principles (Sami HM, 2021).

In this research, we will try to provide effective locational benefits for selling mangoes in Dhaka city. Due to high price provision, increased demand, and multiple selling locations, Dhaka remains an important trading point for the entire Bangladesh, hence research should consider customer locations as well (Kabir & Parolin, 2012). Considering Ks Nearest neighboring algorithm we can figure out the best possible classes for our requirement (Sami HM, 2021). Through this process, we can easily identify the bestselling locations with the highest possible customers by which we can achieve the wastage reduction of mangoes. To observe the forecasted mango growths and demand we could have applied various methods like LSTM and other regression processes (Sami et al., 2021), but due to yearly incremental suggestion by BBS 2018, we have factored a suggested growth of 2.5% yearly in average for increase and price of mangoes. Moreover, the efficiency attained by LSTM for prediction is highly accurate which could have been undertaken as well (Hasan and Arifuzzaman, 2021). 

In this research, we have attributes like distance between production places and selling locations as factors for mango supply optimization. Grimes & AiBDTen have shown that when supply is less the price becomes more similarly with an increase in supply the price drops (Grimes & AiBDTen, 2010). Our research hypothesis is to provide an optimized supply of mango based on the provided price and supply for major mango varieties. In this research, the additionally available mangoes would be distributed for a lesser price to major demand locations following a hypothetical model. We would then observe the similarity of such optimization with PRAN to effectively sell products for cheaper prices (Azim A, 2016). To represent the addressing issues related to efficient mango distribution to provide effective prices both for poor or inferior producers and for lower-income customers, we need to formulate an effective method. The following research would indicate the research gaps and areas for exploration in the Table below: 

Table 1: Research Gap Analysis.

Theoretical Framework

In this research, we will associate pioneering knowledge of gravity location modelling1 and K Means Clustering2 using our available data sources. These two models will help us to answer the research queries formed through available research. These research gaps and original limitations of practical problems have helped us determine, how the mangoes that are currently being wasted could be effectively distributed across the supply chain for more profit for producers and less pricey for customers. 

Research Corpus 

In this research, we have collected the prices of four major mangoes which occupy nearly 67% of entire mango markets from BBS 2021-22. Other mangoes although occupying a significant portion still need to be recorded in the BBS repository. The dataset comprises the prices of mangoes in each year starting from 2014 till 2022. Based on an average observation, during the mango season, it is observed that these four different mango types are ranged through different prices based on different time zones. 

Also, we have accumulated a general approximate production amount of each of these four mangoes in different places provided as metric tons from the BBS 2021-22 data source for mango production. There are around 9 different time segments where the price of mango fluctuates based on supply and market demand. Based on the information available and based on the economic principle of supply and demand, we are planning to provide an optimized mango distribution that will help reduce mango prices and also will help increase profitability for both customers and producers.

This research is carried out to forecast an effective production amount for mangoes in Bangladesh. Its found that using price-based suggestions, we will primarily indicate effective production amounts. After determining the production amounts in different regions, we will initiate k means clustering to segregate different regions using both supply and demand parameters. Finally proposing a hypothetical experiment, we will try and reduce multiple stages of mango distribution that will effectively reduce the mango prices and would provide better profits for producers and more utility for customers. For this research, we have collected data about major population concentrations of Bangladesh. We found that despite the overall selling concentration, the majority of selling is concentrated towards these locales (Mondal et al., 2010). Its also found despite constituting only less than 9% overall land mass these areas concentrate more than 20 million people with more earning and spending capability (Mondal et al., 2010). So, primarily we have segregated the land mass into various population concentration zones using K Means clustering. Based on this we observed that three different high population clusters are generated based on population concentration. 

In this research we have found out that nearly 23 high population zones occupying less than 9% land mass of Bangladesh have about 21.11 million people as urban population. Orsini et al. have found that in 3rd world countries majority of purchasing power remains with the urban population (Orsini et al., 2013). 

In the case of Bangladesh major cities largely occupies that power. After determining the major areas for selling mangoes we have reduced the price volatility of prices using the principle of supply and price. According to Scrimali, when the supply is limited, the price definitely grows higher (Scrimali, 2012). 

Fig. 1: Application of K Means Clustering – Determining varying sales concentration areas.

A similar has happened in the Mango market of Bangladesh. According to the following Fig. 2, we observe that during specific times major mangoes of Bangladesh observe price fluctuations despite being a peak production season in the country.

According to Fig.1, we observe that the prices are given as an average price between different observed sources. But in general, at the very middle the prices become the very lowest and at the beginning and ending the prices are highest. 

Fig. 2: Mango Price at Various Periods during Production Season.

It is also seen that despite being the same fruit, the price standard deviation varies at 7.9 BDT for each Kg. In such a situation a majority of customers who are unwilling to pay high price purchases less than they desire which causes a serious drift of loss. As its a perishable product we propose to produce a less volatile price variant solution that will be helpful for both the customers and producers keeping the same average price intact. Our hypothetical experiment is based on the supply vs price relationship. We have found the prices of four main mango types present in the Bangladesh market. These mangoes have a wide range of demands because of their tastes, color, flavor, and usability. The BBS 2021-22 consumer index dataset has been used to support our research. The research methodology that proposes the price reduction is shown in Fig. 3. The experiment to reduce the price variance and thus propose prices with low volatility and low range is also explained through mango type Gopalbhog as an experiment below in Table 2. In this research, we have found that Gopalbhog has a yearly demand of 920000 metric tons. The four types of mangoes that we discussed in this research have around 21 lac metric tons of production among the whole countrys 25 lac tons of production. We have found out that there is an average price of BDT 61.667 for each kg of Gopalbhog of mangoes. Based on the each below-average production period as indicated by Fig. 3, an excess price exists in the market. To reduce that excess amount of price, an additional amount of production should be conducted during a high price period. 

Fig. 3: Experiment for Price Adjustment.

After the end of higher production adjustment, it is observed that the exact production amount is matched with the total yearly production for Gopalbhog. Similarly, a price range is proposed for each price instance which has a much lower standard deviation of BDT 4.42 for each kg throughout the entire production season. This proposition is much more effective than the price volatility of BDT 8.1 and the price range of BDT 26 for each Kg of mango. In these newly proposed prices for each kg, the overall price range throughout the season is only BDT 15 at maximum. This ensures a better production consumption and better profit commitment for producers and consumers. 

Table 2: Gopalbhog: Production Optimization & Price Proposition.

As we observe from Table 2, we see that throughout the entire mango production season the maximum price is BDT 67.5/kg and lowest price is BDT 55/Kg. If we allocate the prices with a standard deviation of BDT 5 from each instance it will help to associate the following issues.

1. Production Planning to ensure better distribution optimization 
2. Lower price range will ensure more customers 

Production Allocation based on Market Requirement

Currently the price indication of mangoes throughout the season indicates an imperfection in the production process. It shows that based on the current market conditions of provisioning mangoes in the market, the fruit gets a very short period for price dip. In such a short period a sharp price dip does not ensure a large-scale consumption. This is one of the primary reasons behind the wastage of raw mangoes. Our proposed method has shown from Table 2, that how current prices disperse production amounts. In Table 3 below, we are showing a better indication of price-based production and mango supply distribution.

We observe from the above Table 3, that despite Bangladesh achieving a higher target in production and export of Mango in each year, Mango still causes serious degree of parity in comparison to production requirement based on low price volatile demand. 

Table 3: Gopalbhog Production Parity.

Its seen that in each different price instance, there are nearly 2500 tons over or more than 5000 under production happening. Despite the production amount remaining the same for fulfilling the local and export needs, the high degree of price volatility thus disrupts the capability of making mango available for all buyers. This factor is relevant to more favorable economic concepts of inflation rate, purchasing power parity, dollar surge rate, unwavering interest rate, etc. In the following Table 4, we observe the slope-based increase or decrease in demand with mango consumption. It shows how in each price instance, due to a larger price rise or price drop more people will be inclined to either purchase mango or not purchase mango. 

Table 4: SPC for Current Production & Proposed Production Amount for Gopalbhog.

In the confidence interval of 10%, we find that the proposed production amount makes a z-score result with acceptability for more than 9 price instances out of 10 price instances. Moreover, we find that with the same confidence interval of 10%, we find that under current production distribution, nearly 6 price instances out of 10 have been rejected. It means that by statistical process control (SPC) of a 10% confidence interval either some price instances are producing more or less in more than 50% of cases. It certainly indicates that the price of mango remains more volatile in the current condition rather than in the proposed condition. One more observant phenomenon that supports our proposed production solution is that during the price instance that exhibits the lowest price for mangoes certainly remains one of the shortest periods for price change. It suggests that during that period due to ample supply in the market with low prices higher number of customers make greater than their usual amount of purchase (in the case of all the major mangoes it remains from June 25 till July 10 or July 10 till July 20). With the sharp completion of the hugely available supply, the price certainly starts moving back up.  This process makes the following impact –

1. Due to short periodic price drop mangoes gets consumed by more customers who can avail of the fruit with lower prices. 
2. The price drop is directly proportional to the availability of the product because of capitalistic market nature. 

Purchase Power Parity (PPP) & Segmentation of Customer

Its observed that in Bangladesh, in most cases the purchase of Mangoes is associated with a minimum of 5 kg or more. It is also associated with the fact that most purchasers are purchasing for the consumption of their family. Roy & Rahman (Roy & Rahman, 2022) mentioned that in a household survey, most people want to purchase bulk mangoes from the same source rather than purchasing them multiple times. In such a situation it is observed from the current price trends that in most cases the prices are rather moving back up faster creating a very small drop. In our proposed position of production and distribution, the price situations will have a much lengthier low-value period and will also have more effective prices that could be helpful for more mango consumers. In Fig. 4, we have explained this issue with more effective clarification. 

Fig. 4: Standard Deviation of Price/Kg for Proposed & Current Mango Price.

In Fig. 4, we observe that with our proposed method the prices will have nearly half the standard deviation in prices. Also, in our proposed method we are maintaining the consumption of the same number of mangoes but with lower price deviation the customers are expecting better prices. Similarly, this also leads to lower price rises that will help more consumers to make an effective purchasing decision. Also, we find that Roy & Rahman explanation that in case of a price rise people tend to purchase or consume a smaller number of mangoes. Similarly in the current situation of price with more standard deviation in the price, it shows that there is high variance in price which supports the purchase ability for more customers. Hence in the current price condition, the following two things are happening:

1. With the sharp price increment, fewer customers are becoming interested in consuming more mangoes. 
2. With sharp price rises lesser number of mangoes are getting consumed because consumers are expecting a price drop that will allow the bulk purchase. 

Considering the similar principles by Negi & Wood, when perishable goods do not get proper customer concern at the right market time then their actual worth starts getting diminished (Negi & Wood, 2019). Similarly, Krishnan et al. have suggested an effective and efficient distribution process that supports environmental sustainability and proper resource usage and recovery (Krishnan et al., 2020). Our research process thus followed a similar qualitative approach but only kept Mango as a major fruit defined as a perishable good.  Following the well-known machine learning model named K Means clustering which has been successfully applied to segregate qualified financial assets from underqualified ones (HM Sami et al., 2022); we have decided to make separations of price segments that will segregate the prices for different customer segments of mangoes to reduce wastage and increase or support more sales among general mass. Although the selection of various variability factors such as peoples purchasing power, type of mango choice by customers, availability of mango during the period, production and availability in the market, price of mango, taste, and flavor of mangoes, usability as secondary products all should be evaluated through naïve bayes algorithm process to understand customer criteria and type for purchase and their purchase quantity in a better manner (Sami et al., 2021); but for this research we are considering specific price factor only for consideration. 

K Means Clustering – for selection between Current Prices & Proposed Prices 

Its found that in Table 5, the current prices with higher standard deviation propose more high price clusters (total 4) and also same number of middle range prices, whereas the proposed prices hold less high price clusters and more middle range price clusters that is 3 and 5 in numbers respectively. Although by proposed price we didnt find any additional price clusters for lower segments. But we got confined by the requirement to fulfill the complete level of production distributed through the quality price ranges. But through our proposed method we can observe that people will be able to get more mangoes for more lower price ranges. Also, in our proposed research we have demonstrated that the highs of the current mango price are way higher than the highs of the proposed price. The overall demonstration of the clustering result is shown below in Table 5. 

Table 5: Cluster Proposition for Current Vs Proposed Price.

Scopes of Improvement & Limitations 

This research has been carried out under the pretense of BBS price data source which is not the exact price source for mango selling price. The mangoes throughout the entire country are sold in various ways, hence the price remains extremely variable under different conditions. Considering the areas of improvement, we have discussed several issues below: 

1. Price of mangoes differs valiantly based on the method of procurement and purchase contract. Hence our research lacks prices for minority purchasers. 
2. This research suggests production amount and price provision as major concerns without evaluating the necessary conditions for the follow-up of such measures 
3. This research process didnt suggest how the supply chain bottlenecks like five different hand exchanges before final consumer reception could be reduced for better price provision for the customers. 
4. Mostly this research lacked quality data support to suggest quality knowledge creation to propose better supply chain optimization.

Through ML methods and SPC, this research has proposed a process that can effectively increase customers benefits to make consumption better by offering lower prices for each mango unit. Similarly, it proposes a lower high price (BDT 67/kg) in comparison to the current price (BDT 71/Kg). Moreover, the price distribution is highly effective as it allocates more meaningful 5 middle price range clusters in comparison to the current 4 price range clusters. In terms of high price range clusters, our method proposes three high cluster ranges in comparison to the currently proposed four price range clusters with higher high prices. Mostly we find that in terms of production distribution support by SPC, nearly 6 price segments were rejected by current production allocation. We find that in the case of proposed price ranges only 1 price segment has been rejected and that is also due to low production. Hence, we can suggest that our proposed research methods are well supported for better Mango distribution and allocation both in terms of better price suitability and also for better price variance. 

H.M.S.: Conceptualization, visualization and contribution to investigation; K.A.A.: Investigation, writing and checking the manuscript, editing and funding acquisition; S.H.C.: Writing and checking the manuscript along with editing; P.N.R.: Writing and checking the manuscript along with editing; K.A.D.: Writing and checking the manuscript along with editing; H.M.: Writing and checking the manuscript along with editing.

The authors would like to thank the Correspondence, Hasan M Sami, Senior Lecturer, School of Business and Economics, North South University, for his support in this research along with the scholarly writings, research articles, and blogs that helped in writing this present paper, which are duly referred to relevant places of this article as well as in the reference list below.

All authors declare no conflict of research for this study.