Obviously, we are always looking ahead. When Googling “the best quotes about the future,” you are taken to a predictable listicle of the top 10 aggregated future quotes that say something like “Study the past if you want to define the future. - Confucious” I couldn’t think of a better quote when it comes to data predictions. While data can make everyone an oracle, not everyone knows the capability of creative artificial intelligence.
Going back to our first post, we stated the goal of this blog was to inspire creativity in our users by providing them with thoughtful posts on creativity, the future of work, and data literacy in relation to machine learning. In other words, when creating content, we think about the user’s decision-making process and how they can be creative with machine learning to improve it.
Let’s talk about AI’s relationship with creativity for a second.
Can AI Be Creative?
Perhaps the better question would be “Can you be creative with AI?” If you think of AI as a tool you collaborate with rather than a job replacement technology, the answer is YES. You can be creative with AI and improve your workflow with it.
A lot of us use Google’s Smart Compose in our emails and even when you log into your G Suite, you are presented with suggested documents based on your previous data. This is just scratching the surface into adding machine learning into the business user’s workflow. AI is subtracting mundane tasks from the workflow and leaving time for creativity. AI can present you with the best suggestion based on your data, but at the end of the day, it’s the human being creative with the technology.
Here are some examples of how humans have been creative with AI:
- YACHT created a whole album with Machine Learning using MIDI data
- The Seattle-based augmented writing company, Textio, is using AI to make more inclusive job listings.
- The design company, AI Build, uses AI to reduce material and labor costs to build nearly impossible structures.
If you want to hear more about artificial creativity, here’s a video I like on using AI in the creative process:
An Explanation of AI, ML, and Predictive Analytics in Non-Technical Language.
So, if you’ve made it this far you’ve noticed I’ve thrown some terms at you without defining them and using them seemingly interchangeably. Despite popular, non-technical belief Machine Learning is not the same as Artificial Intelligence and Machine Learning and Artificial Intelligence is not the same thing as Predictive Analytics.
Ugh, I hate to be this guy, but the easiest example to use is from the film Ex Machina.
In Ex Machina, the robot Ava (Alicia Vikander) is the AI. Oscar Isaac’s character is the CEO of a giant search engine company. He unethically scrapes the data from all his fictional user’s search queries and plugs the data into a database and programs an algorithm (or several) causing Ava to act as human-like as possible based on the data. This is Machine Learning.
Here are the definitions of AI, ML, and Data Predictions in non-technical language:
Artificial Intelligence: A non-human system that shows human-like intelligence. AI is an umbrella term, which includes machine learning and other techniques. ML is AI but AI is not ML. Examples include playing the computer at a video game or Siri.
Machine Learning: A method to teach an AI system to perform a task based on the data given to them. Basically, you are "training" a "model" using ML techniques to make predictions. (Often this model is also called an ML model). Examples include Waze giving you the best route home based on user traffic data.
So What Are Data Predictions?
Using Machine Learning with AI can create logical solutions to questions. In the most basic sense, you can program an AI system to give you predictions using Machine Learning. AI is the system, Machine Learning is the method, and a Data Prediction is the task.
Here's a deeper dive into the defintions of AI, ML, and Predictive Analytics.
To Make Creative Predictions, You First Need Data.
And the more you have, the better decisions you can make and more creative you can be. There are many ways to go about collecting data and using it to make predictions.
- Collect data from surveys
- Observe and record data
- Interview subjects
- Conduct a focus group
- Use online databases such as Kaggle or Google datasets
Here’s an awesome post I found all about using Kaggle for your data science initiatives.
Isolate a Business Problem and Aim to Solve It
Data can be qualitative (numbers) or quantitative (attributes) and easily be inaccurate depending on your data collection procedures.
Before collecting data, I recommend isolating a business problem, defining it, and choose a metric you want to better understand. I also recommend getting the maximum information out of a few variables as possible. You don’t want your prediction model to use unnecessary variables and ruin the trend you’re trying to clarify.
Data cleaning is important to remove any possible errors and delete duplicate entries. If you use the Obviously AI platform, we make sure your data is squeaky clean before it predicts an outcome.
Humans Can Create Biased Data
One of the biggest flaws in humans collecting data is unconscious biases which could affect the overall prediction you're trying to make. Before you try to use machine learning to make predictions with your data, peer review your findings and discuss your thought process behind collecting data.
AI is only as powerful as you allow it to be. To truly be successful, you need accurate data to make accurate predictions. You can’t be Confucius if your data is wrong.
How to Creatively Predict Based on Data.
The value of predictions is limitless. Smarter business decisions are made every day using AI technology, creating a competitive edge for those who prioritize data collection. Now business users can predict customer churn, credit card fraud, labor costs, demand, daily foot traffic—oh you’re still reading—maintenance, competitor analysis, financial risks, etc. using historical data.
Like I mentioned before, Machine Learning is an Artificial Intelligence method to solve a problem. At the heart of Machine Learning is algorithms that build predictive models.
What are Machine Learning Predictive Models?
The terms “Machine Learning” and “Predictive Modeling” can be used almost interchangeably. Data analysts use Machine Learning to build predictive models to show the most likely possible outcome.
What is Predictive Analytics?
Predictive analytics is the analysis of data to see patterns and relationships between variables to identify future outcomes.
Now that we have defined our key terms, let’s show a little bit of what machine learning can do for your business.
Going back to Obviously AI’s blog introductory post, I highlighted an example of how to be creative with data by using climbing and weather data from Mt. Rainier.
I downloaded the data from Kaggle into a CSV file, looked it over, and decided I wanted to ask a question in plain English.
What causes high success rates when summiting Mt. Rainier?
I asked this somewhat selfishly because if I did choose to summit the active volcano for the first time, I would do it successfully based on data and look like an all-telling alpinist oracle when passing by unlucky climbers who chose to descend.
The top driver was (somewhat obviously) the route you chose to take getting up the volcano. This example is valuable in a number of ways—not just to measure summit success rate.
From a business perspective, with this data you can also:
- Predict what products climbers need to ensure maximum success rate and market to them accordingly.
- Build content around the top drivers of summit success i.e route and gear guides.
- Predict sales and a perfect price point for a product based on seasonal weather conditions.
Additionally, some other data you can gather to make your predictions more powerful include success rate depending on if the climber had certain products (crampons, ice axes etc.), when guided tours are most valuable or necessary, forecast demand ahead of the climbing season. These are just some of the possibilities.
And this example doesn’t even scratch the surface of what is possible creatively with ML.
Creative Machine Learning Can Do Things Like Reveal Potential Profit.
Airbnb hosts are running a business. They need to know how to price their homes for incoming guests. Imagine you are brand new to hosting and have no idea what the value per night of your entire home in Manhattan is. If you price it too low, you could miss out on profit. If you price it too high, you miss out on business. You can take a quick look around Airbnb’s site to make an educated guess on where to set the price or you can use Machine Learning.
Using the data set, I predicted the price using attributing factors like neighborhood group and room type. Using the data, you can conclude the top driver by impact is the room type.
The average predicted price for the entire home is $211.38/night. From the price is the highest in the neighborhood group of Manhattan.
This is a good starting point for a novice Airbnb host, but there are other factors at play here in determining your potential profit.
Imagine you wanted to go deeper and ask your data more questions.
Let’s take a look at average price vs. number of reviews:
The price generally trends down if you have more reviews, indicating your Airbnb profit will decrease as time goes on.
From the Kaggle data, you can also average out the minimum nights in the dataset. I already know the answer, so I’ll just tell you it’s 7.02 nights. Multiply average minimum nights by average price and you’ll make ~$1,483.88/week with the possibility of the profit decreasing as reviews start rolling in. Additionally, don’t forget to subtract Airbnb’s pay cut, which is typically under 13%.
There Are Many More Examples of Creative Approaches to Machine Learning.
If you want to see more examples on the possibilities of what you can achieve with data predictions in business. To see more possibilities, visit our case studies:
- Calculate dynamic pricing
- Reduce churn
- Detect fraud
- Predict mitigation risk
- Find subrogation possibilities
- Predict customer behavior
- Optimize assortments
- Improve operations and marketing
- Plan labor
- Predict customer lifetime value
- Predict net promoter score
- Model marketing funnel
- Find the best ways to cross-sell and up-sell
- Attribute multi-channel marketing
- Predict default rate and financial crime
- Manage cash
- Predict loan demand
- Predict gamer churn
- Predict gamer purchases
- Understand live stream stats
- Optimize gameplay
- Predict wins and losses
- Predict KPIs
- Predict in-app purchases
- Build customer personas
- Reduce affiliate/referral fraud
- Reduce readmissions
- Predict drug adherence
- Predict no shows
- Predict length of stay
- Predict churn
As our blog grows and matures, we will go more into these use cases to show exactly how businesses use ML to creatively do these tasks.
Creativity Will Set You Apart in the Machine Learning Field.
The interesting thing about predictive analytics and modeling is you don’t need the same technical skills as you needed before. Think of technical skills as hardware and the basic concepts of machine learning as software. It’s much harder to change the hardware than the software. It’s time to start thinking of machine learning as a creative tool, rather than inaccessible technology reserved for techies.
Use machine learning to see the future for yourself. So, what would you like to predict today?