Online shopping refers to a delivery and pick-up service. This particular service is offered by companies through a website and mobile app. The service allows customers to order items from participating retailers with the shopping being done by a personal shopper or company. Covid-19 has changed the way the world operates. The ‘new normal’ has also seen dramatic increases in online shopping trends for apparent reasons. Contactless transactions are safer, and lockdowns have given people no choice but to use online platforms to purchase goods. As you know, however, don’t last forever. Still, once a customer has experienced the convenience and safety of an initially obligated online shopping experience, they are very likely to use it again even when they don’t have to. Some of the advantages of online shopping include; there are a variety of products, it's covenient, there are a lot of online discounts, there is buyer/shopping secrecy, the customers can avoid crowds, you can find unique/rare products and there is detailed information about the products that you are buying. To effect some of these advantages of online shopping a recommendation system or recommendation engine can be used for information filtering where it tries to predict the preferences of a user and provide suggestions based on these preferences.
To provide a recommendation system that will help both existing and new customers. Our main of creating a recommendation system is to categorise the customers based on their purchase quantities and set this as our guide in recommending the products based on the size of a business.
-
To particularly provide product recommendations and to look for unique ways to personalize marketing to customer base.
-
To determine the month with the highest sales so as to plan the optimal time to give discounts.
-
To determine the peak hours of the day inorder to better strategize our marketing through online ads.
-
To determine if the unit price affects the quantity demanded
Our data set is from UCI MachineLearning Repository. It is from an non-store online shop selling unique gifts. The Data set collected is from 2010 December to 2011 December.
The data has 8 features;
- InvoiceNo
- StockCode
- Description
- Quantity
- nvoiceDate
- UnitPrice
- CustomerID
- Country
Collaborative Filtering
We used Cosine Similarity to recommend various Products to customers.
Memory_Based
We used the SVD model to create our model and test our data.
RMSE
Memory based We used KNN (K-Nearest Neighbours) With Means model has an test RMSE value of 0.244 and cross validation RMSE value of 0.246.
Model based
We used the SVD (Singular Value Decomposition) model has a test RMSE score of 0.364 and cross validation (CV) RMSE score of 0.363.
Each client will have a variety of products suggested to them as they are gathered filling out missing entries in the matrix during matrix factorization using SVD.
Special thanks to our Moringa School Data science Techincal Mentors for their guidance throughout the process of our project and the amazing Elites team members 👇