Track B: E-commerce retail merchant demand forecast and inventory optimization problem
Problem background:
There are thousands of merchants on the e-commerce platform. They will place the goods in the warehouse supporting the e-commerce platform, and the e-commerce platform will conduct unified management of these goods. Through scientific management methods and intelligent decision-making, the supply chain driven by big data intelligence can significantly reduce inventory costs while ensuring on-time performance of goods. Generally speaking, the above supply chain optimization problem will include the following aspects:
There is a historical shipment list of e-commerce retail merchants (Appendix 1), which gives the historical shipments of each merchant in the past 6 months. The daily shipment volume of goods stored in different e-commerce warehouses. Assume that the shipment volume is the historical demand for each commodity in each warehouse. At the same time, you can also get information about each product, merchant, and warehouse (Appendix 2-4), such as classification, brand, effective date, etc. The selection and introduction of this information will help better predict and manage inventory in the supply chain.
Preliminary questions: (continuously updated)
First of all, part b itself also divides the problems, one is demand forecasting, then the second part is the optimization of warehouse inventory, and the third part is to consider a demand under special circumstances.
Question 1: Use the data in Appendix 1-4 to predict the demand for each merchant’s products in each warehouse
from 2023-05- 16 to 2023-05-30. Please fill in the prediction results in Result Table 1 and upload it to the competition platform, and evaluate the prediction performance of your model. Please also discuss: Based on the data analysis and modeling process, how can these time series formed by merchants, warehouses, and commodities be classified so that the demand characteristics of the same category are most similar?
Idea:
First of all, let’s talk about the first question. First of all, we need to use the data in Appendix 1-4 to predict the demand, and then conduct a performance evaluation after the prediction, and at the same time examine how to make these categories better in terms of demand characteristics. most similar.
Let's first observe the data form. First, this is Table 1. Then the table relies on product name and product number to connect to Table 2. At the same time, Table 1 relies on seller number to connect to Table 3, and then uses Warehouse number to connect to Table 4. Come connect. Then we want to merge table 1234. We want to connect table 1234 to form a large table. Then the most intuitive thing to think of is to use the PD merge function in Python. Then we will show it in the code later and show it to everyone. After you merge it, the table will become a large table like this. Here is We want to predict the label, and then the rest are its features, and then after processing the data, I make a prediction.
The first step is to merge the tables, and then preprocessing. First, you observe whether the data is zero or has obviously wrong values. Then you conduct a correlation analysis, that is, use this function to observe its correlation. Then output the highly relevant features as features for subsequent prediction. If we mainly examine this prediction over time, then we can use Arima, which has typical time series prediction, or use LSTM. Of course, the simplest sentence is linear regression, which can be adopted. Then we provide an evaluation of the performance limit. We can output a picture of the predicted result graph and the actual value-added graph, or output our accuracy, and then ask whether it is good, that is, how do we do it? categories, you can use them as similar categories.
and the previous pair basically use cluster analysis. Of course, we will give a more detailed introduction later. We will introduce it to you based on the code and the result icon. Okay, let’s analyze a problem in detail first. , and then we will expand on questions 2 and 3 later, and everyone is welcome to pay attention, and we will continue to share them.
Question 2: There are some new merchants + warehouses + product dimensions (Appendix 5). The reason for this may be newly launched products, or some products have been changed. The warehouse where it is stored. Please discuss how these newly emerged prediction dimensions can be referenced through the data in historical attachment 1 to find similar sequences and complete the prediction values of these dimensions from 2023-05-16 to 2023-05-30. Please fill in the prediction results in Result Table 2 and upload it to the competition platform.
Idea:
Question 3: Large-scale promotions occur regularly in June every year, which brings great challenges to accurate demand forecasting and contract fulfillment. Attachment 6 gives the demand data for the merchant + warehouse + product dimension corresponding to Attachment 1 during last year’s Double Eleven. Please refer to these data to give the forecast values from 2023-06-01 to 2023-06-20. Please fill in the prediction results in Result Table 3 and upload it to the competition platform.
Idea:
This time, we will provide the complete solution ideas and codes for Question B, and also share some paper templates and other materials. Friends who need it can pay attention and continue to update! The complete solution code can be obtained by clicking here
# https://
# mbd.pub/o/bread/mbd-ZZWalpty