Recommendation System
Copyright: © 2023
|
Pages: 16
DOI: 10.4018/978-1-6684-4730-7.ch009
Abstract
Recommendation systems are critical tools used by marketing departments to provide customers with product recommendations. Data scientists also use recommendation system analysis to assess the effectiveness of product and service suggestions. There are two types of recommendation systems: content-based and collaborative filtering. Content-based recommendations are based on the customer's purchase history, while collaborative filtering suggests a product based on purchasing behavior. Collaborative filtering can be divided into content-based filtering, which suggests products based on similar purchasing behavior, and item-based filtering, which suggests products based on their attributes. User-based and item-based nearest-neighbor collaborative filtering and probabilistic methods are used to analyze data and provide product suggestions. Businesses rely on recommendation systems to achieve various objectives such as customer retention and increased ROI. RapidMiner can analyze data following the principles of the recommendation system, as demonstrated in this chapter, step-by-step.
TopRecommendation System
The recommendation system is divided into two types.
-
1.
Content-Based Recommendation: It focuses on making recommendations using past content as a reference (Tai, Sun & Yao 2021). For example, when a customer has previously purchased Product A, the system will present Product A and other products that are similar in properties to Product A.
-
2.
Collaborative Filtering: CF focuses on analyzing customer behavior (Emon et al, 2021: Khatter et al, 2021). For example, Customer A tends to buy soap and shampoo all together, so Customer B who buys soap will likely buy shampoo too. The aforementioned system is also divided into 2 forms.
-
a.
Content-Based Filtering: This is to focus on analyzing user behavior such as a group of customers with similar purchasing behavior (Yadalam et al., 2020: Kamukwamba & Chunxiao, 2021).
-
b.
Item-Based Filtering: This is to focus on analyzing product properties (Kusumawardhani, Nasrun & Setianingsih, 2019: Lourenco & Varde 2020: Padhy, Singh & Vetrivelan, 2022). For example, scooter and skateboards are different products but share similar properties.
The CF processing algorithm is a prediction of the similar data from the existing data, as in the example.
Table 1.
Examples of datasets for forecasting data with recommendation system
|
Data Mining Techniques |
Python |
Design Thinking |
Technology Management |
| Sarawut Ramjan |
5 |
2 |
5 |
4 |
| Jirapon Sunkpho |
2 |
5 |
? |
3 |
| Kom Campiranon |
2 |
2 |
4 |
2 |
| Purimprach Saengkaew |
5 |
1 |
5 |
? |
Complete Chapter List
Search this Book: