Build a tool to set up Marketing campaigns on a product line or service pack. This tool automatically collects comments from Facebook fan pages and groups according to each campaign. The collected data is analyzed and used to predict the positive and negative feedback from users about these products/services.

Machine learning algorithms and computational infrastructure that can be applied are:

• NLP with Bigdata technologies.
• Deep learning to predict user opinions.
• Visualization and statistics about user views.
• Web / Mobile interface.

Refers to web/app search experiences that are tailored specifically to an individual's interests by incorporating information about the individual beyond the specific query provided. There are two general approaches to personalizing search results, involving modifying the user's query and re-ranking search results. And build recommendation for searches based on historical information about transactions, amounts, and information about usage habits.

Dinogo is an online travel application and hotel is our top prioritized focus in 2020, with thousands of hotels can be booked for every destination that can be an overwhelming task for customers. In order to provide the best experience for customers in hotel booking, the recommendation system is a must, that is to predict and recommend five hotels to a user that he is more likely to book in a hundred distinct hotels. To support this, Dinogo has a dataset of customer behavior historical data, e.g hotels that a particular customer has searched/viewed/booked in the past.
Note: Dinogo’s dataset will be provided to teams doing this challenge after they have signed a NDA (non-disclosure agreement - which states that the dataset must NOT be used for any purposes other than doing this challenge).

With an enormous database of more than 60 million mobile subscriptions, our ambitious goal is to develop a digital bank for Vietnamese - ViettelPay. To achieve this, Viettel Digital is motivated to conduct research on credit scoring. To clarify, a credit scoring model is a tool that is typically used in the decision-making procedure of accepting or rejecting a loan.
The model is operated based on a statistical/machine learning system which processes and analyzes borrowers’ credit information in order to distinguish between "good" and "bad" loans and estimate the default probability. By utilizing our substantial source of telecommunications data, we are striving to build a forecasting model that is able to automatically offer each applicant’s credit score which is human-readable and easy-to-interpret.

Develop Social Listening application to collect and analyze product’s price for Viet Nam real estate market.

Build AI Bot to assist customers, users in many areas of life. It can be able to:

• Apply voice recognition
• Analyze natural language
• Understand the request and able to response to user's queries
• Able to execute actions following user's request
• Artificial Intelligence can learn from its interactions and become progressively intelligent.

There are hundreds of thousands of new ads posted on Chotot on a daily basis that being viewed by millions of potential buyers everyday. These are items ranging across different categories such as Vehicle, Property, Electronics, Pets and others hobbies and appliances. How to make full use of these fresh contents data to create an application or web feature that increase the personalized user experience in buy and sell such that one person will find joy, fun and good deals everyday from the first day visiting Chotot.

Expected deliverables: Demo application based on anonymized data provided by Cho Tot.

FEC is currently offering many products and services over the digital channels and thus face many online frauds from customers such as using fake IDs to apply or using unauthorized information from others to apply for loans online.

Suggested Solution: Develop solution to detect fraudulent documents submitted by customers such as National ID, Driver License, Motor Registration Certificate, Family book...Develop solution to raise red flags to potential frauds using mobile data or any other data such as social data.

Expected deliverables: An interface built on Facebook Messenger and Zalo where customers can apply for loans/credit cards or perform loan/credit card requests

Dynamic Pricing: Automate adjust the products's price base on sales volume. Once the sellers have a lot of SKUs, pricing control will be become very difficult, the automatic update pricring base on the pre-configured rules makes it easy for the seller to control the sales of a product, help them easily optimizing profits, and selling more products.

An MShop is a business that accepts payments with MoMo. It could be anything from a supermarket like Coop Mart, to a convenient store like Circle K, to a restaurant like Mon Hue or Thai Express, to a coffee shop like Phuc Long, to a street food vendor like MotoCoffee. There are more than ten thousand MShops all over Vietnam, cover the majority of a person's daily needs. For users, they usually want to explore a new service or try something new with confidence. However, there is usually too many options to choose from. We want to make it more convenient for users to find something new, based on merchant and MoMo data. A few directions that we can do: • Make it easy for users to search businesses within a location. They could even see some of the reviews about the shop or the business.

• Recommend MShop based on user's history and preferences (personalization).
• Offer targeted promotion for users. MoMo will provide anonymized user data including their transactions, and some other information for building the app and the personalization model.

On an e-commerce website like Sendo.vn, there are millions of customers and millions of products. In order to provide a customer with a great shopping experience, it is important to understand his or her preferences and recommend to him / her products that are most likely of his / her interest. By providing personalized product recommendations, the website will also greatly improve its revenues. Given Sendo's proprietary dataset of historical customer behavior data (e.g., the products that a particular customer has viewed in the past), how would you build an effective product recommendation system using machine learning?

Note: Sendo's dataset will be provided to teams doing this challenge after they have signed a non-disclosure agreement (which states that the dataset must NOT be used for any purposes other than doing this challenge).