Before getting into the topic, I wish to ask you one question. Just imagine if you are gone for an outstation trip and suddenly if you feel hungry, there are two restaurants in front of you, remember you don’t know the local language and you don’t have a mobile with you. Now say based on what you find a better restaurant between them?. Obviously, you choose the best one based on the customers who prefer right?
Our topic is also brought from a similar kind of idea only. But you know the reality? Now it is ruling the tech world particularly the e-commerce sector. Okay, let’s get into the topic.
What is Affinity Analysis?
Affinity Analysis is the kind of predictive analysis technique that does the process of data mining and fetches the hiding insightful correlation between the different variables based on their co-occurrence happening in between the individuals or the groups in the dataset. This Affinity Analysis actually gives us valuable information by grabbing the knowledge from the relatively similar entities to make the decisions wiser like the recommendations to the users.
How does it work with Big Data?
In fact, Affinity Analysis gains a lot of attention while it reveals the hidden patterns and relationships that take part inside the massive amount of datasets, called Big Data. Those interesting insights in the sets of data are uncovered by the affinity analysis with the help of Association Rule Mining.
The procedure of Association rule mining involves finding the frequency in the attributes and it tests such attributes to check whether they satisfy certain prefixed criteria with necessary support and confidence. In this process, it finds the features with similarities (say conditions). It numbered the similarities in the co-occurrence present in the data set. In simple words, if the co-occurrence feature (X) with the predefined condition is present then the other feature (Y) with a certain similarity exists.
In Association Rule Mining the first feature (X) is known as Antecedent and the co-occurrence feature with considerable relationship (Y) and further features called Consequents. These are defined based on the factors that support and confidence in Association Rule
The factor Support in Association rule mining is defined based on the popularity of the feature sets. And it is used to measure the strength of the relationship between the features.
For example: In the 10 combinations of the given dataset product A is listed in 5 combinations. Hence the support factor of Product A is 5/10.
The confidence range is between 0 to 1. In the 10 combinations if we find the confidence value of product A is equal to 1, then we conclude that product A is taking part in all the 10 combinations.
Where to use Affinity Analysis
The concept of Affinity analysis almost employs in all the sectors. The primary goal behind the usage of this type of analysis is “Maximising the benefit with minimizing the effort”. In this article, we discuss the two classic sectors which have more usage of Affinity Analysis, namely
Application of Affinity Analysis in Clinical diagnosis
The medical records of the patients obtained from clinical trials play a vital role in creating association rules in the dataset. The information gathered using the association rules helps to find valuable insights for doing effective diagnosis in the medications. Similarly by analyzing the association between the recorded symptoms and the chance of having a disease preventive measures are taken into consideration to avoid such disease attacks. And also in the pharmaceutical sector based on the respective trials and the association found in curing the respective disease every particular drug is getting approval for the market.
Application of Affinity Analysis in e-Commerce
This is the most notable sector in the effective usage of Affinity Analysis. In the e-commerce industry, a particular terminology is very much popular and the name of the respective terminology is known as “Market Basket Analysis”. It’s actually not a new process, the affinity analysis is only referred to as the Market Basket Analysis in the e-commerce sector.
This Analysis tries to find the pattern between the seller and the customer in general to increase the selling strength of the particular seller or the selling agency. Initially, this Market Basket Analysis originated in the retail sector. Later the e-commerce industry grabs and uses this particular analysis type in a much more effective manner with the help of technical tools like Excel, Python, Google Analytics, Tableau, etc., and much more technological stuff.
In e-commerce, the first record the transaction details of the customers in detail. Based on that data they further classified the customers with similar kinds of preference and based on the purchasing history of a person or group they recommend similar kinds of items to the other person or groups having similar kind of preference of the previous person/group. Initially, they use Affinity analysis findings as to the recommendation to similar customers. This helps the sellers to “increase their profit with minimizing the cost” by identifying where to promote which items to increase the selling potential.
Example for your better understanding
Let’s imagine a retail shop, the shopkeeper wants to gain more profit that’s why he/she recorded the customer’s past purchasing data. Based on the recorded data he/she classifies the customers into several categories. Such categories are classified based on the customer’s transactions in his/her shop. In that transaction data, they make categories based on the itemsets like “customer1=[carrot, milk, almond, potato]”, “customer2=[tea powder, milk, lemon, sugar]”, and so on.
Now based on the similarity in the item lists of the customers he/she finds the association between them. If customer A purchases the products (1,2,3), then based on the association similar customer B is found, and if B purchases products (1,2,4), then the chance is more than customer B will purchase product 3 if he/she got recommended to purchase that product.
The media shown in this article are not owned by Analytics Vidhya and is used at the Author’s discretion.