Problem type: Customer Segmentation / Clustering
Finding new customers is vital in every industry. The process for finding new customers begins by learning as much as possible from the existing customers. Understanding current customers allow organizations to identify groups of customers that have different product interests, different market participation, or different response to marketing efforts.
Market segmentation, the process of identifying customers’ groups, makes use of geographic, demographic, psychographic, and behavioral characteristics of customers. By understanding the differences between the different segments, organizations can make better strategic choices about opportunities, product definition, positioning, promotions, pricing, and target marketing.
- Exploratory Data Analysis - EDA - in order to identify the most valuable variables to create patterns.
- Clusters development:Using SOM on top of Hierachical clustering.
- Relevance of findings with a business application.
- Presentation of the project to WWWW company board members.
Problem type: Machine Learning - Predictive Model
In the hotel industry, as in many other travel-related industries, demand is managed through advanced bookings. Bookings (also known as reservations) are a forward contract between the hotel and the customer that gives the customer the right to use the service in the future at a settled price, but often with an option to cancel [^1]. This cancellation option puts the risk on hotels who have to honor the bookings that they have on-the-books, but, at the same time, have to support the opportunity costs of having vacant rooms, when someone cancels, and there is no time to try to sell the room or sell it at a discounted price. In Europe, the cancellation rate by reservation value, from 2014 to 2018, rose from 33% to 40%.
Hotel chain C, a chain with resort and city hotels in Portugal, isn't any different than other independent and non-independent hotel chains. Hotel chain C was severely impacted by cancellations, representing almost 28% in H1 and almost 42% in H2.
To reduce the uncertainty about demand, the director wants to implement prediction models to allow the chain’s hotels to forecast net demand based on reservations on-the-books. With these models' estimations, the director expects to implement better pricing and overbooking policies and identify bookings with high likelihood of canceling. Identifying those bookings could allow the hotels to try to contact those bookings’ customers and make offers to try to prevent cancellation (e.g., dinner, car parking, spa treatments, discounts, or other perks). Michael's goal is to reduce cancellations to a rate of 20%.
- Exploratory Data Analysis - EDA - in order to identify the most valuable variables to create patterns.
- Modeling of the solution using the Random Forest algorithm.
- Deployment approach.
- Presentation of the project to Hotel Chain C company board members.
Problem type: Data Mining - Segmentation
This dataset was provided by the Paralyzed Veterans of America (PVA). PVA is a non-profit organization that provides programs and services for US veterans with spinal cord injuries or disease. With an in-house database of 13 million donors, PVA is also one of the largest direct mail fundraisers in the United States of America.
As an non-profit organization, PVA needs donor´s contribution and it is necessary to provide to each donor the causes in which he is more interested, in order to feel affinity. One group that is of particular interest to PVA is "Lapsed" donors. These are individuals who made their last donation to PVA 13 to 24 months ago. They represent an important group to PVA, since the longer someone goes without donating, the less likely they will be to give again. Therefore, recapturing these former donors is a critical aspect of PVA's fundraising efforts.
- Exploratory Data Analysis - EDA - in order to identify the most valuable variables to create patterns.
- Segmentation development: using the SOM on top of hierarchical algorithm in order to create clusters.
- Interpretation of each cluster.
Problem type: Customer Segmentation / Clustering
Finding new customers is vital in every industry. The process for finding new customers begins by learning as much as possible from the existing customers. Understanding current customers allow organizations to identify groups of customers that have different product interests, different market participation, or different response to marketing efforts.
Market segmentation, the process of identifying customers’ groups, makes use of geographic, demographic, psychographic, and behavioral characteristics of customers. By understanding the differences between the different segments, organizations can make better strategic choices about opportunities, product definition, positioning, promotions, pricing, and target marketing.
- Exploratory Data Analysis - EDA - in order to identify the most valuable variables to create patterns.
- Clusters development:Using SOM on top of Hierachical clustering.
- Relevance of findings with a business application.
- Presentation of the project to WWWW company board members.
Problem type:* Business Intelligence - Support Decision Making
My client, "Javi Indian foods Ltd" , is the owner of two indian food takeaway restaurants. The first restaurant, located in London, UK is open since 2015 and motivated by the success of the business, Javi bought another indian restaurant in Tampa, USA in 2017, that restaurant has been operating also since 2015 and also worked only in a takeaway concept.
In order to improve his business and so increase the sales, our client requested our services of business intelligence, to support his further decisions.
The menus differ a little from one restaurant to another. In London, our client realized that his target clients, tourists, really appreciate mainstream indian dishes. Since Tampa is a more residential place, the clients demand more diversity there, this way the menu of the second restaurant has more dishes. Both restaurants sell drinks, sauces and food including small snacks.
Sales Forecast:
- How much orders will we have next month?
- What is the affluence throughout the day/ week/ month/ year?
Inventory Management:
- What are the most/least ordered dishes in each restaurant?
- What are the most ordered categories in each restaurant?
Marketing:
- What is the average price per order?
Human Resource Management:
- How many orders is an employee in charge of?
- How is my staff productivity?
Creation of a data warehouse with a dimensional model that translates the different business entities relations.
- How can I store and monitor the business data?
Problem type:* Data Mining - Recommend Items
ManyGiftsUK is a UK-based non-store online retailer that is focused on selling unique all-occasion gifts. The company was established in 1981 but has only recently shifted completely online. With the new data the company has collected, they expect to build a recommender system that is able to facilitate user choices. Additionally, a particular challenge is the cold start problem - how can we suggest relevant items to new customers?
- Exploratory Data Analysis - EDA - in order to identify some patterns and insights on the data.
- Modeling, using Alternating Least Squares - ALS.
- Evaluation of the final model using, namely, precision and AUC.