From the course: Learning Data Science: Understanding the Basics
Comb techniques for predictive analytics
From the course: Learning Data Science: Understanding the Basics
Comb techniques for predictive analytics
- So far, everything you've seen has been about the past. You've seen how to collect different data types, then perform statistical analysis. These statistics were a starting point to greater insights. Your data science team can create correlations between certain events. Now, let's take these ideas and turn them around. We'll take your insights and flip them around to predict the future. This is typically called predictive analytics. This term is closely associated with data science. Sometimes they're even used interchangeably, but the two are not the same thing. Predictive analytics are a type of data science. Remember that data science is applying the scientific method to your data. Predictive analytics takes the results and makes it actionable. Think of it this way. Meteorology is a type of science. It studies physics, wind speed, and the atmosphere. If you're outside with a meteorologist, they'll show you how the clouds look a certain way and how the pressure determines their movement. That's the science of meteorology. It's about understanding the weather and seeing historical trends. Yet, most people don't talk about meteorology like a science. Instead, they ask about weather forecasting. Weather forecasting is when a team of meteorologists use predictive analytics. They can use probability and correlation to predict weather patterns. Meteorologists can use historical data to assign probabilities. There also might be a correlation between low pressure systems and severe storms. As the pressure decreases, the severity of the storm increases. There's a positive correlation between pressure and storms. All this analysis comes together so that the meteorologist can answer a simple question, "What's the weather going to be like tomorrow?" What was about understanding the past, now becomes a prediction for the future. There's a new interest in predictive analytics. It's because new tools and technology allow for more interesting insights. Think about weather forecasting. Right now, the Weather Service is restricted to historical data from just a few thousand stations. Now, imagine that the Weather Service gave out millions of beacons. People would install them in their homes and hook them up to their wireless networks. These inexpensive devices would record air pressure and temperature information. They would also record video and audio. Then they would upload their data to a Hadoop cluster. This would give these scientists unprecedented levels of information. That's why predictive analytics is so closely associated with big data in data science. The higher volumes of data allow the team to ask more interesting questions. Then the team can perform complex analysis. Here, the team would be able to look at the weather patterns house by house and block by block. They could create complex predictive models based on millions of homes. The same holds true with your team. Think about our running shoe website. Imagine that your team collects millions of Tweets about running. The team identifies a few influential runners. You could send them free shoes or promotions, hoping they would say nice things about your company. You could also use this data to identify key events. These new tools allow you to take a much larger view of your data. Your team can analyze millions of Tweets, just like the Weather Service can analyze petabytes of information. Then you can look at waves of information in real time. A general rule to keep in mind is that the more data you have, the more powerful and accurate your predictive analytics will be. Organizations usually get really excited about the idea of predictive analytics. So much so, that they don't always put enough time into developing their data science teams. They want to go straight to predictions without understanding their data. When you're working on the data science team, be sure to communicate that the quality of the predictions depends on how well the team has analyzed the data. Your team has to better understand the past to be able to predict the future. Don't short change your analysis. Ask good questions of your data, and use your statistical tools to create interesting reports. Once you do that, your predictions of the future are far more likely to be accurate.