# I. Introduction ith increased globalization and advancement in technology, the retail market has become more and more dynamic, and therefore, retailers need a new approach to identify different objectives to be more competitive and successful. Inventory management is one of those key sectors that determine the success of a retailer. In today's ever-changing climate with a high level of uncertainty, keeping up with the demands leads to positive result on the market. Mining or extracting customer insight from structured and unstructured data and other sources is of tremendous importance for inventory management in retail stores. The change in customers' taste plays a significant part of what product is to be stored. Predicting which product will give more profit, products that are sold in unison, information like that is useful to store products in the inventory. Knowing which that product is out of fashion can help us in optimizing an inventory effectively. Some of the popular data mining techniques area) Clustering b) Association rules c) Decision tree Data mining is finding and predicting hidden information from databases. It is a powerful technology with great potential to help organizations focus on the most accurate data in their data warehouses [1,2,3]. Every technique can play its part in decision making and storing products, some more than others. The analysis from each of these techniques produces patterns which helps us to find valuable information. For a successful business, identification of high-profit, lowrisk customers, retaining those customers, and bring future customers are important tasks for business owners and marketers. # II. Related Work In the last few years, the internet gives us new business concepts and also much information. As competitive pressure rises, the application of data mining process in customer's behavior becomes a excellent tool. [6] Customer relationship management (CRM) aims at stronger loyalty of customers with feasible market share. With competition for shelf space intensifying, there is a pressing need to provide shoppers with a highly differentiated value proposition through "right product mix in the right amount at the right time." [7] Customer relationship management (CRM) and customer profiles: Federated department stores are combining customer and transaction data to identify the best customers and offer exclusive extras. [8] # III. Methodology and Analysis a) Clustering The inventory space in a retail store is a precious commodity. To represent products seasonality, retail stores need to organize the products with care. Festivals and holidays should also be kept in mind when reforming the shelf. If a product has a large amount of sales in a day, it shouldn't dominate over other product storage. The storage priority is given to a product that has a high sale rate for an extended period. To represent products from a retail store we have used this set as an example-The data set given above is a series of data set representing the amount sells of each of the product weeklies. To get useful information out of this data set, we use a simple clustering technique, which is k-means clustering. # K-means clustering K means defines a prototype in terms of a centroid, which usually the mean of a group of points and it is used for objects in a continuous n-dimensional space. Centroid never corresponds to an actual data point. Basic K-means algorithm-1. Select k points as the initial centroid 2. Repeat 3. For k number of clusters by assigning each point closest to its centroid 4. Recomputed the centroid of each cluster 5. Until centroids do not change. To apply k-means clustering to our data set we have used software known as Rapid Miner studio. To reduce the dominance of a product after one day of massive amount of sell over our inventory, we normalize the dataset. After normalizing the data set, we get-After applying k-means algorithm, we get- # And the centroid table- In our dataset, the optimal number of k=6 from the performance vector. In this dataset, the average sell is -1 and higher the disparity from -1 the larger or smaller the amount of sell. We can see that for milk in week one, the amount of sell=-0.384 and for week two is 5.170. The highest disparity from all the weeks is week two that suggests the high amount of sell. If this is the 1st week of November, then there is a high chance to sell this time next year, so for future storage, we can use this information and store a high amount of milk or a high amount of frozen items for the 1st week of November. As more milk gets sold, it should also give us the idea of which product will be out of stock first. It will also help to apply FIFO(oldest stock gets sold first). That means the product that to be out of stock early can be sold first. We can make an early prediction that milk to out of stock next year's November first and store milk as quickly as possible. # b) Association rule Association rule mining analysis is used to find patterns that suggests how strongly associated features in the dataset. Implication rules represent these patterns [4]. Finding the most useful role in and collecting interesting patterns to improve the organization of storing products is one of our main goals, and association rule will help us in that regard. The popular algorithms that use association rules include AIS, SETM, Apriori, and variations of the latter. As can be seen, from support normalized products 2 and 3 are sold together at a 60 percent rate. That tells us to store normalized products 2 and 3 together to increase efficiency. Perhaps a highly-priced normalized 2 product can be stored beside normalized product 3 to ensure a maximum profit. Now applying association rule algorithm, we get-Confidence is the conditional probability of an event if given a set event has occurred. From this, if someone already bought products 6, 5 and 1, the conditional probability of someone buying product number 2 is .827, which is the highest from this group of data sets. As can be seen, product number 2 should be stored close to 6 or 5 or 1 to increase efficiency and selling. Lift suggests the randomness of the given rule. A positive value which is more than 1 suggests how reliable the rule is. From the dataset, we can see that association rule number 46 is the most useful rule. branch suggests an outcome of a test, and each leaf node holds a class label [5]. Here the algorithm came into the conclusion that gender is the root node. The decision tree tells us that, age group is the key element while storing for female customer. Age group of more than 89.5 are most likely to be loyal and under 89.5, we check churn and other attributes that tells us which one is in need of prioritization. # IV. Conclusion As the retail industry gets ever so competitive, it is necessary for us to find every single opportunity to have the edge over everyone. Inventory management plays a major part of retail industry, and data mining techniques can be of use to store products efficiently with the future in mind. Customer insight is essential for any department even in storing products, and with these data mining techniques, valuable information can be extracted and used to our advantage. Our goal is to increase the attention in inventory management with the help of these techniques as it gets overlooked. ![Support ({X} ? {Y}) = ??????????? ?????????ð??" ???? ? ??? ? ????? ?????? ?ð??" ???????????? Confidence ({X} ? {?}) = ???????????? ?????????ð??" ???? ? ??? ? ???????????? ?????????ð??" ? Lift ({X} ? {?}) = (???????????? ?????????ð??" ???? ? ??? ?) / (???????????? ?????????ð??" ?) ???????? ?ð??" ???????????? ?????????ð??" ?](image-2.png "") ![](image-3.png "") ![](image-4.png "") ![](image-5.png "") ![](image-6.png "") ![](image-7.png "") ![](image-8.png "") ![](image-9.png "") © 2020 Global Journals © 2020 Global Journals Application and Analysis of Retail Inventory Using Data Mining Techniques ## And after applying the FP-growth algorithm Here we can see a performance measurement unit called support. It tells us the frequency of different or individual items occurs together. * Data mining techniques for marketing, sales and customer support MJ ABerry GLinoff 1997 John Wiley and Sons USA * Advances in Knowledge Discovery and Data Mining UFayyad GPiatetsky-Shapiro PSmyth RUthurusamy 1996 AAAI Press Menlo Park, Calif * JiaweiHan MichelineKamber Data Mining Concepts and Techniques MorganKauffman 2006 2nd ed * Association Rules in Data Mining: An Application on a Clothing and Accessory Specialty Store * Efficient Classification of Data Using Decision Tree Bhaskar N. Patel, Satish G. Prajapati and Dr. Kamaljit I. Lakhtaria * Data Mining for the category management in theretail market Jochen Garcke, Michael Griebel and Michael Thess * Data Mining for Retail Inventory Management Pradip Kumar bala * Growing Lifetime Value, Chain store Age(75) May 1999