Purpose To create a model that will provide a daily forecast of gross weight of shipments expected to be processed by a given Shipping Company over a period of 2 months over the period of 2020-06-14 to 2020-08-13. This is a time-series forecasting challenge. The created time-series model should be able to forecast the daily gross weight of shipments using 16 months of historical data between 2019-02-14 and 2020-06-13. The gross weight of shipments will help in negotiating costs for a large number of shipments. Please note that the target variable gross_weight is not grouped by day in the train file (train_3_4_pr.csv), but it can be easily calculated. Just group shipments by date and sum gross weights. It will be your target variable. Input Dataset • train_3_4_pr.csv - File that contains historical shipments over a 16-month period between 2019-02-14 and 2020-06-13. This file can be used to predict the number and weight of shipments for each Shipping Company in the following 2-month period between 2020-06-14 and 2020-08-13. More than one record in the file might have the same 'shipment_id', but 'shipping_company' for such rows are different. 'shipping_company' reflects potential shipping companies which might be used, and the 'selected' column shows which company (in some cases more than one) has been selected. Please note that in this part of the competition (#3) train file contains only 'selected' companies meaning that 'selected' is always equal to 'Y'. # shipment_id - Shipment ID # send_timestamp - Date when the order was sent to the destination country (in 'source_country' timezone, e.g. UK - GMT) # pick_up_point - Pick Up Point abbreviation # drop_off_point - Drop Off Point abbreviation # source_country - Country from which the goods need to be shipped # destination_country - Country to where the goods need to be shipped # freight_cost - Cost of transportation/k # gross_weight - Gross weight in kg that needs to be shipped # shipment_charges - Fixed cost per shipment # shipment_mode - Method of shipment (e.g. air, ocean) # shipping_company - Candidate shipping company # selected - Whether the company in 'shipping_company' was selected or not • shipping_companies_details_3.csv - File that contains all necessary information about shipping companies # source_country - Country from which the goods need to be shipped # destination_country - Country to where the goods need to be shipped # shipment_mode - Method of shipment (e.g. air, ocean) # cut_off_time - The time window in which goods can be picked up (e.g. "10 - 2 and 3 - 6 BST" means that pick up is available between the times of 10 a.m. and 2 p.m., and again between the ties of 3 p.m. and 6 p.m). All times are in Bangladesh Standard Time. # tat - Turnaround Time is the time it takes from a package order to shipment. T refers to the time a package was sent, and CO means cut-off time, or the time after which a package can no longer be shipped. "Before CO – T+0" and "After CO – T+1" means that if the goods are sent to the shipping company before the cut off time, then the package will be sent on the same day. If the package arrives to the shipping company after the cut off time, then it will be shipped out on the following day. # processing_days - Days on which Pick up Point works # pick_up_point - Pick Up Point abbreviation # drop_off_point - Drop Off Point abbreviation # min_sc - Min Cost of Shipment # max_sc - Max Cost of Shipment # shipping_company - Name of shipping company # shipment_charges - Fixed cost per shipment Submission File • submission_4.csv - File that will be submitted by the participants. The file SHOULD NOT CONTAIN ANY (date+shipping company) ID OR FIELD TITLES; the order of the target variables should be according to the submission file format but the first row and first column should be removed prior to submission. The platform will only accept csv files that consist only of numeric values. # ID - Date in the mm/dd/yyyy format + Name of the shipping company (should be removed before the submission) # weight - Gross weight of shipments (target variable). This column is empty and should be filled out by the participants. Evaluation Metric Submissions are evaluated on e^(-RMSE/MAX(observed_values)) where Root Mean Squared Error is calculated between the predicted values and observed values. *While calculating the final ranking, the scores will be normalized (separately for each competition) to eliminate any inherent bias in the evaluation metric for the problem statements. We will normalize the scores before applying the formula (20% of Challenge 1) + (40% of Challenge 2) + (40% of Challenge 3). The score of each competition will have a valuable impact on the final score since they will be in the same range.