Instacart Market Basket Data Analysis - Reorder Analysis
The skills the author demoed here can be learned through taking Data Science with Machine Learning bootcamp with NYC Data Science Academy.
Introduction
InstaCart market basket data analysis was a Kaggle competition that was open early 2016 and was conducted by Instacart. Instacart, a grocery ordering and delivery app, aims to make it easy to fill refrigerator and pantry with personal favorites and staples when needed. After selecting products through the Instacart app, personal shoppers review the order and do the in-store shopping and delivery for customers.
Instacart uses transactional data to develop models that predict which products a user will buy again, try for the first time, or add to their cart next during a session. The objective for this competition was to use anonymized data on customer orders over time to predict which previously purchased products will be in a user’s next order.
Data was provided as separate datasets with specific entities such as orders, products, reorders, etc. The datasets had to be combined appropriately for data analysis and visualization as well as for modeling. Data provided had no missing or null values. The training and test set were combined in a single dataset and had to be separated into different datasets. The training set contained 131209 rows, the test set contained 75000 and a prior set containing reorders contained over 3 million records. Due to computing restrains, this dataset was reduced to 300K rows for modeling.
Exploratory Data Analysis (EDA)
-
Number of products in each department
The following plot helps us understand how Instacart has stocked products by department. Based on the plot, the most number of products is available under the personal care department. Followed by snacks and pantry.
2. Number of products in each aisle
The following plot helps us understand how Instacart has stocked products by aisle. Based on the plot, the most number of products is available under the candy chocolates. Followed by ice cream and vitamin supplements. This contradicts with the number of products by department. However, let's analyze further.
3. Number of products based on department and aisles
The following plots divides the number of products among aisles and department. The highest number of products in personal care is vitamin supplements which was also one of the highest number of products based on aisles. Similarly candy chocolates is the highest under the department Snacks. So now we find a relationship between the number of products in department and in aisles.
4. Number of reorders in training set and priorset
There is more reorders in the prior set which is quite obvious considering the numbers of rows in the prior set
5. Reorder for each day in training set and prior set
The reorders in both dataset are quite same. The highest reorder happening on day 0 which we assume is a Sunday, followed by a Monday and then on a Saturday
Data Modeling
Using Apriori algorithm the following set of items were obtained that were a possibility of being reordered by itself or with other items. Due to constraints of computing and time, the entire analysis could not get further than this. With more time and computing available, I would like to perform more algorithms and more analysis on the models obtained to ensure the model is close to accurate.
About Author
