The specific problem in this project is about the time-series data trend prediction. The specific application scenario is in e-commerce. You are given a real dataset obtained from a real-world e-commerce application where there were 1000 products and 31490 customers (i.e., buyers) who bought these products. Of these 1000 products there are 100 key products (popular products). Also these 1000 products are in 15 categories. The specific data are given in the seven tables and the specific details of these tables are given below. The time window of this dataset is in 119 days with data documentation for each day. Hence, the time unit is one day where the timeline goes from the 0-th day to the 118-th day (17 weeks in total). Now you are asked to do the sale quantity prediction for the 100 key products for each day between the 119-th day and the 146-th day (four weeks).
• buyer_basic_info.txt: the basic attribute information of the buyers; in particular, the column names of this table are "buyer_id", "registration_time", "seller_level", "buyer_level", "age", and "gender". If we do not know the gender of a buyer, we set this buyer's gender attribute as -1.
• buyer_historical_category15_quantity.txt: the consumption quantities in the 15 categories for the buyers; in particular, the column names of this table are "buyer_id", "consumption quantity in the 1st category", ..., and "consumption quantity in the 15th category". The 15 categories are the ones of the products the customers bought in this dataset.
• buyer_historical_category15_money.txt: the consumption amounts in the 15 categories for the buyers; in particular, the column names of this table are "buyer_id", "consumption amount in the 1st category", ..., and "consumption amount in the 15th category".
• product_features.txt: the basic attribute information of the products; in particular, the column names of this table are "product_id", "attribute_1", "attribute_2", and "original price".
• Key_product_IDs.txt: the key product IDs
• trade_info_training.txt: the trade information between the key products and the buyers from the 0-th day to the 118-th day (17 weeks); in particular, the column names of this table are "product_id", "buyer_id", "trade_time", "trade_quantity", and "trade_price".
• product_distribution_training_set.txt: there are 120 columns, where the 1-st column shows the "product_id" and the 2-nd to the 120-th columns show the "quantities" of the key products from the 0-th day to the 118-th day; for example, the element at the 5-th row and the 10-th column in this table shows the quantity of the 5-th product at the 8-th day.
For grad students you are asked to do the prediction for the overall sale quantity of the 100 key products for each day of the four weeks (i.e., for each of the time window from the 119-th day to the 146-th day), and also for each key product for each day of the four weeks.
This phase is for the coding part of the project and concerns with the implementation of a time-series prediction method that you either take from the literature or you have developed by yourself as the result of your research in the first phase.
Please make sure to follow the format requirement as the text output file specified here. The file puts each prediction as one line where the first prediction is for the overall prediction and each subsequent prediction is for a key product. Each prediction output line begins with the key product id where the overall prediction id is 0. There is a space between the prediction and the key product id. Then there is a space between a pair of the predictions of two neighboring days. The prediction lines in the output file begin with the first line as the overall prediction where the product id is 0, and then the first key product prediction with the smallest product id (i.e., 1), all the way to the last line as the prediction for the last key product prediction (i.e., id = 964). Also note that for undergrad students your output file only has one line prediction just for the overall prediction beginning with the product id = 0.
What you need to turn in: you shall turn in a zipped package containing the source code of your implementation of the prediction method with appropriate comments and documentations in the code, a README file to explain how to compile and run your code under what specific environment, and a text file containing the output matrix following exactly the format requirement stated above.
Attachment:- Data.rar