Simulation of Agri-Retail Store and Association Mining

Simulation of Agri-Retail Store and Association Mining

Abstract

Today simulation software’s are used in the real-world systems for smooth functioning and better results of the general model in on which the system is already being run. This paper works on the simulation of an agricultural based retail store for each of the three seasons, to find out the system usage at each point in the system and to optimize the number of staff members that are required to run the system perfectly without extra usage of expenses and keeping the customer wait times adequate. This would also help in avoiding over staffing and understaffing.  In addition to that association rule mining is done to predict the customer behavior. This helps in customer pattern of buying the products using the market basket analysis in the retail store.

Keywords

Agri-Retail Store, Simulation, Association Rule Mining (ARM), Seasons.

Introduction

An agricultural product-based retail store goes through three seasons throughout the year due to its dependence on the weather, rainwater and other climatic conditions. The three basic seasons are summer, winter and rainy. The rainy season sees the highest number of customers due to the abundance of water which is essential for farming. The summer season sees the least number of customers and the winter season sees the customers between the intermediate range of the two seasons. A retail store generally faces the problems like customers having to deal with longer wait times, lack of face to face time of the staff with customers which ultimately leads to loss of customers reducing the total turnover of the business.

For this paper, the case study of a retail store located in one of the towns in India is taken. As most of the farmers are uneducated, it is important for the staff members to give face to face time in addition to the time which is required to serve the customer. In this store, the customer either goes for a packed product or an unpacked product. The products range from various types of seeds to pesticides and herbicides. The year begins with the rainy season, which is from June to September, then comes the winter season which is from October to January and the third season is the summer season which is February to May. By optimizing the number of workers, considering that the store runs at full potential, the maximation of profit for the store is done in the paper.

Get Help With Your Essay
If you need assistance with writing your essay, our professional essay writing service is here to help!
Essay Writing Service

Later the association mining of the rainy reason is done as it sees the highest number of customers, also all the type of products are sold in the rainy season in order to find out the association by which the farmers go for pesticide or herbicide after choosing a specific type of seed. Association mining here is required to find out the support, confidence and lift if the customer buys maize or bajra or groundnut and then goes for herbicide or pesticide. This will have help in better providing deals on the products bought repeatedly, better store alignment and other marketing schemes to successfully thrive in the increasing number of retailer’s day by day.

Literature Review

In this project the modelling and simulation procedure for in-store retail shop was produced to organize and examine customer and stock flows. One of the main objectives of this paper was to propose an approach for the building of a simulation model using Arena Software primarily analyzing the customer congestion and utilization of the cashier register data. This proposed procedure was applied to the actual case and a simulation based on the further literature review was conducted to examine the customer flow showing the usefulness of the same to improve the cashier and worker scheduling. The first paper suggests [1] the use of arena software for a similar modelling and simulation of aluminum brake brackets and to the formulate future planning process to increase efficiency and costs while decreasing the labor work and identifying the bottlenecks of the process. Using the help of the process flow chart using the Input Analyzer tool in Arena the software data was fed into it and simulated while using other optimizing tools such as the opt Quest and Process Analyzer was used. With a replication time given as 30 days a total of replications were done where the results were analyzed it was found out that major changes could be done by just manipulating resources available and keeping the process the same with minor differences. With just the optimization of the total plant layout various factors such as wait time, machine idle and wait times also transportation and docking time were reduced significantly. Another paper proposes a simulation-based modelling for the online retailers promotional pricing [2]. As of today, online retailing is the most important supply and distribution channel with respect to the ever-increasing consumer demand and has been expanding vastly since the last decade ever since. With rapid development and increase in the information technology consumers can order whatever they want with various choices to choose from safely and reliably with more and more increase in online shopping. Thus, retailers offer various kinds of promotions to increase the flow rate in the system. Here to in the following paper Arena and opt Quest were used to simulate and analyze the model and a sensitivity analyses were performed to evaluate the impact of the stock inventory costs acquired to unit shipping cost to learn how to optimize it more and increasing profits. We learnt from this literature review that various assumptions are required to be made and many simulations with different assumptions give us results and outputs based on what kind of a procedure is to be followed for a problem. Even with limited data the best fitting purchase amount, conversion and

stock inventory rates can be calculated. The relaxation of different assumptions develops more and more optimized results and might result in complicated problems. Another paper classifies three general types of classes [3] into manufacturing system design, operation and simulation package

development. The authors reviewed over 290 papers giving us a comprehensive idea about the topic giving us a literature of defined research trend we need to follow while simulating a problem. It extends the classification schemes giving us an idea about the data mining approaches and modelling processes. The results were then the basis of the conclusion on which we can base our future work which is covered in the project. With a wide range of complex systems simulation has become useful in a spectrum of fields not limited to manufacturing, healthcare, supply chain, cellular, aerospace, military and hundreds of other applications. Discrete event simulation is one of the most commonly used techniques in identifying the dynamics of a system. Thus, the development using optimization tools such as SIMIO, Arena and incorporating them into real world problems are truly beneficial to enhance the performance of the packages and for the betterment of the manufacturing systems. There will always be a need for better efficient techniques for the growing complexity of the manufacturing applications. Supply chain management plays a vital role in the problem dedicated in this project. The inventory management aspect in the undertaken problem helps us make the process more efficient on optimizing it using

the simulation processes. Lowering inventory cost and conducting a steady supply and demand chain helps increase the profits in the process. This indirectly helps in the demand forecasting and in the management and availability of labor work to cut costs down to the minimum. Optimization techniques help us in drawing insight from a process system of complex nature with various constraints and factors for coverage of demand and high volumes of data.

Data collected

In this study, the data to be collected is to be analyzed in the input analyzer of the arena software to find out the specific distributions or each of the data sets. This will help in finding out the equations for each of the distribution to run the simulation. There is also a data set collected for association mining which is done with the help of python. The data collected here is:

Customer arrival times

The time customer takes to decide if he wants packed or unpacked products

Time taken by the staff to serve customer for unpacked and packed products

Time taken for payment at the cashier counter by the customer

Association between the purchase of the products

The real time data is collected from the store itself recording all the times for the working hours from 9.00 am to 8.00 pm each day of the three seasons. The total of 11 working hours for each of the staff members. This model runs on some of the assumptions like there are no breaks between the shifts. Also Sunday is also considered as working day.

Now the simulation is run for 120 replications considering each month has 30 days. All the values of time are represented in minutes.

3.1 Data analysis

The data is put in the input analyzer to study the various distributions of the various data sets of the three seasons.

Fig 1. Interarrival time

Distribution: Beta

Expression: -0.5 + 13 * BETA (1.57, 3.02)

Fig 2. Time taken at the cashier

Distribution: Gamma

Expression: 0.5 + GAMM (0.312, 5.35)

Fig 1. and Fig 2. shows the distribution and the equation which represent the distribution for rainy season. The first figure represents the beta distribution  for inter arrival times.

Fig 3: The simulation model of the retail store

 

The equations are then input in the arena model to find the service time, wait time, total time in the system, system utilization and usage. Fig 3. shows the arena model of the retail store. The sequence of events in the flow diagram of the model is as follows:

The customer enters the store, the arrival time is recorded, customer places the order on the counter i.e., the customer checks for the availability of the required product, the type of order is places, packed or unpacked product is then served to the customer, customer count for packed and unpacked products is done, the customer goes to the cashier counter, the customer exits the system.

Simulation

Simulation is done using the arena software. Small scale businesses generally ignore simulation tools due to the lack of resources or information and the mentality with which they are built. The benefits of simulation are not only useful in running of these businesses smoothly but also leads to better overall customer satisfaction.

Association Rule Mining (ARM)

It is a data mining technique used to find the relationship among the purchased products and to discover unpredictable results for the purchase databases.

Apriori Algorithm

It is an algorithm useful in finding the association between the frequently repeating items in the database. This further helps in finding the general trends of such items. Some basic concepts like support which is the popularity of the item, confidence, which is likelihood of buying an item if a certain item is bought.

In our model it is used to find the customer product buying patterns. The data sets consist of the items customer buys and the items bought proceeding to it. The association rules in the program are built on minimum support 0.0045, minimum confidence 0.2, minimum lift 3 and minimum length 2. The file is imported as the location of the file is input in the program, the parameters given above are set and the program is run.

Results

7.1 Simulation

After each of the three simulations are run, on the actual condition of the retail store, it was found out that there is scope to decrease the staff members. For rainy season it is found out that decrease in 1-1 staff person at packed and unpacked product counter does not increase the wait times much. For summer and winter season it is found out that the model can run smoothly even if we reduce two staff members.

 

Fig 4. Result table for simulation of three seasons

 

With actual condition for rainy, summer and winter season the average service time was 5.0433, 5.6265, 6.5859 minutes, the average wait time was 0.8731, 0.4415, 0.3213 minutes and total time in the system came to 5.9174, 6.0680, 6.9073 minutes respectively. After decreasing the staff members these values came to 5.682, 1.0141 and 6.0822 respectively. One more simulation for rainy season is also done considering the minimum staff condition which is using 1 staff member at each part of the system. The results are still acceptable, but the appropriate condition would be to prefer lesser wait times due to the unpredictable scenarios that can happen during the real-world trading. Fig 4. Shows the result table for the simulation of three seasons with actual condition and optimum condition.

7.2 System Usage and Utilization

The system usage and utilization are important characteristics to study the system so that the there are no unnecessary expenditures. This will also help us better understand if the system at any point is underutilized or overutilized. The below graphs compare the utilization of the staff members which are represented by blue bar as cashier staff, red as counter staff, yellow as staff for packed products and green represents staff for unpacked products. As seen in actual condition and the condition that is run after the simulation after decreasing 1 staff member each from staff for packed and unpacked products, the graphs show some increase in the utilization without letting the system go to extreme utilization at any point.

Fig 5. Graph comparison of system usage of rainy season

Fig 6. Graphs showing utilization and usage for winter season

 

Now, the Fig 5. and Fig 6. represent the system utilization and usage for winter season after the simulation is done. Each color here too represents the same staff as above.

7.3 Association mining

The results for association mining for only the rainy season are discussed as the model would smoothly function for both other seasons due to the farmers following almost similar pattern of buying products. It is found out that if customer buys maize, the chance of buying herbicide or pesticide is 100%. If the customer buys bajra, the chance is buying pesticide is 46% while the chance of buying herbicide is 60%. If the customer buys groundnut, then the chance of buying herbicide is 60% while for pesticide its 57.14%. Fig 7 gives the results for support and confidence of seeds to pesticides and herbicides.

Fig 7. Association rules for retail store by Support × Confidence

Conclusion

At the initial working condition at the firm, there was not proper utilization of the manpower. Our simulation model showed us that increasing the number of workers would not result in reduction of the waiting time for each customer, rather result in monetary losses for the firm.

By decreasing one staff member each at the packed and unpacked section, the waiting time is almost still the same and the customers are served equally effectively and therefore increasing the number of customers that can be served daily.

By Using this simulation, understaffing and overstaffing can be avoided and help in saving large sums of money in form of payments to unnecessary staff. This way the firm can invest that money in foundation of new branches or expansion.

Customer buys both herbicide and pesticide after he/she buys maize. After the bajra seeds are bought, the chance of customer buying pesticide or herbicide is less than 50%. The chance of customer buying herbicide or pesticide after buying groundnut is 60 and 57% which means they are needed more than for bajra seeds.

Future Scope

This model is very flexible and can be applied to many such businesses which follow similar trends of customer arrival and inter arrivals with the same system in the operation. With the use of this model the increase and decrease in the number of staff members can lead to better usage of the utilities.

The simulated model of the actual model should be implemented if the resources are to be used for further branch expansion while the actual model should be used if the store itself needs to be expanded internally.

For the firm’s expansion, where there is similar trend of customer arrivals and inter-arrivals with the same system in operation, we can use our current system for its analysis and generate maximum output. But for a different branch with different customer arrival and inter-arrival times a new simulation will be needed to be designed.

This simulation can be used only for walk in order. With further improvement it can be used for online orders. As we can see there are few customers that are not served due to the unavailability of the product, the system can be modified to manage inventory as per the demand.

The simulation could also be set up to receive customer feedback for the service provided and the products purchased for the further improvement of the firm. The System designed is versatile and can be applied to other such businesses working on similar concepts with little to no modification in the system.

The association mining clearly shows the customer product buying trends. Combining of the products which are bought together frequently for schemes and discounts can boost the product sales and help increase the customers. Based on the association rules, the cross-selling strategies such as this can be done by bundling the high value product with less sold product.

References

1)      Neeraj, Nithin, Niranjhan, Sumesh, and Thenarasu. “Modelling and Simulation of Discrete Manufacturing Industry.” Materials Today: Proceedings 5, no. 11 (2018): 24971-4983.

2)      Becerril-Arreola, Leng, and Parlar. “Online Retailers’ Promotional Pricing, Free-shipping Threshold, and Inventory Decisions: A Simulation-based Analysis.” European Journal of Operational Research 230, no. 2 (2013): 272-83.

3)      Negahban, and Smith. “Simulation for Manufacturing System Design and Operation: Literature Review and Analysis.” Journal of Manufacturing Systems 33, no. 2 (2014): 241-61.

4)      Malik, U. (2018, August 09). Association Rule Mining via Apriori Algorithm in Python. Retrieved from https://stackabuse.com/association-rule-mining-via-apriori-algorithm-in-python/

5)      Tupkar, S. V. (2016, January 17). A Simulation Model of Starbucks Cafe using Arena Software. Retrieved from https://www.slideshare.net/sagarvtupkar/a-simulation-model-of-starbucks-cafe-using-arena-software

6)      Figure 2f from: Irimia R, Gottschling M (2016) Taxonomic revision of Rochefortia Sw. (Ehretiaceae, Boraginales). Biodiversity Data Journal 4: E7720. https://doi.org/10.3897/BDJ.4.e7720. (n.d.). doi:10.3897/bdj.4.e7720.figure2f

7)      Moffitt, C. (n.d.). Introduction to Market Basket Analysis in Python. Retrieved from https://pbpython.com/market-basket-analysis.html

8)      Owygs156. (2018, February 16). How to Create Data Visualization for Association Rules in Data Mining. Retrieved from http://intelligentonlinetools.com/blog/2018/02/10/how-to-create-data-visualization-for-association-rules-in-data-mining/

9)      Khader, N., Lashier, A., & Yoon, S. W. (2016). Pharmacy robotic dispensing and planogram analysis using association rule mining with prescription data. Expert Systems with Applications,57, 296-310. doi:10.1016/j.eswa.2016.02.045