Problem 1: Fraud detection among subsidized student meals DETECTION AMONG SUBSIDIZED STUDENT MEALS

Short description of the proposed problem:
The project was designed in cooperation with Margento R&D, which developed a system for subsidizing student meals. When monitoring the data, certain patterns can be observed, that lead to the suspicion of fraud in the use of student coupons. The company wants to understand the fraud scenarios, so it is their goal to develop a fraud detection system and identify patterns in data that would further serve as the basis for incorporating similar technologies into existing solutions.

Short description of data:
The key task of students in the project is creative thinking about possible fraud scenarios, and through the program R recognizes them in the file (.csv), which includes individual transactions. Thus, the student faces a concrete problem and acquires the competences of creative thinking, teamwork and the sense of transferring analytical knowledge from theory to practice.

Name of the company who will provide data: Margento R&D, d.o.o.

Problem 2: Planning of sales territories and delivery routes

Short description of the proposed problem:
The problem considers a synchronized tactical planning of sales territories and corresponding delivery routes for 3PL /4PL providers. Such a planning is challenging in FCMG markets where Traditonal Trade is still strong compared to Key Accounts. In such markets, 3PL/4PL providers have exclusive distribution agreements with several brands. As the rule of thumb, each brand insists on a sales-force that sells only its products. On the other hand, the delivery of products from different brands can be consolidated. The brands are compatible if such consolidation is possible. The lists of the shops of two compatible brands usually overlap to a great extent.

The planner of sales territories for a single brand usually has complete freedom to organize territories for his/her sales-force. A single salesperson thus gets a territory that is often splitinto sub-territories, where each sub-territory can be handled during a single working day. Goods that are sold during a working day are either delivered during the next 24 hours, or the next 48 hours. All sub-territories of the sales-force related to a particular brand and assigned to a particular day can be identified with the daily list of shops in them. Based on the daily lists of shops from all its brands, 3PL/4PL provider plans its delivery routes. Since the planning of sales for different brands is not synchronized, it results in more visits and consequently in more kilometers and more vehicles engaged.

We need a model and an appropriate algorithm for synchronized planning of sales-force sub-territories for compatible brands that result in efficient (less kilometers and less vehicles) delivery routes.

Short description of data:
Data are organized into several tables.
Brands – list of brands with the number of sales people, their working time and upper limit for the number of shops that can be visited during a day;
Consolidation points – depots, cross docks with latitude/longitude;
Vehicles – number of vehicles at consolidation points and their capacity;
Shops – list of shops,  where latitude/longitude, relevant consolidation point and general stop time are specified for each shop;
Visits – list of shops to be visited for each of the Brands, where average time that salesman spends at a shop and average quantity of goods sold is specified for each of the shops.

Name of the company who will provide data: Infora Research Group

Problem 3: Optimization of retail supply chain of a product under stochastic demand

Short description of the proposed problem can be found here: ESGI-SI_Problem3.

Problem 4: Predicting the quantity of meat sold

Short description of the proposed problem:
In 2010, the Panvita Group created the brand AVE, specializing in meat production. The brand comprises of AVE Grill, tasked with preparing grilled dishes, and AVE Chef, tasked with preparing seasoned meat for everyday use.  The company wishes to reduce the quantity of discarded meat and to offer their consumers meat prepared as freshly as possible. For this purpose, the company set a task to predict the quantity of meat sold per day. Currently this job is being carried out by a worker with years of experience in meat procurement organization.
Students are tasked to create a model, which predicts the purchasing of meat for days in advance as accurately as possible.
Operating procedure:
The current operating procedure is as follows:
–    Based on experience and number of orders through the day the worker makes an estimation for the procurement of meat.
–    The procured meat is processed.
–    The prepared meat is supplied to the customers.
The process from procurement to supply takes 2 days. The prepared meat has an expiry date of 8 days.

Short description of data:
Data of daily orders and supply is at the disposal of the company since 2008.

Name of the company who will provide data: Panvita