To make it simple, predictive analytics is the concept of assuming that something that happened in the past will happen in the future. In order to come up with classification of data in predictive analytics we need to stages: the learning and prediction stge. During the learning stage the goal is to teach your model to discover hidden relationships and rules from historical data. After this stage, prediction analytics predict new labels in new data classifications.
A business example can be a person’s level of interest in watches. In order to come up with some data you could apply any of several text-analytics tools that can discover such correlations in an individual’s written text (social network statuses, tweets, blog postings, and such) or online activity (such as online social interactions, photo uploads, and searches)
Predictive analytics is used in our brains on daily basis, as an example in making smart decisions through our past experiences (historical data). Think about the case of waiting to try your clothes in the fitting room.
Parameters to consider:
Expected output: fastest line
Target variable: the shortest amount of time to finish shopping
Historical data: previous shopping experiences
Predictors: number of items, length of line, age of customers, gender etc
Why Predictive Analytics is important for organizations?
Optimizing marketing campaigns. Predictions are mostly used by organization to promote cross-sell opportunities, to understand and determine customer purchases which directly lead growth of profitable partnerships and customers.
Detecting fraud. One of the reasons why predictive analytics is used, is to prevent criminal behavior. Predicting advanced persistent threats an organization can improve pattern detection and prevent any suspicious act.
Improving operations. Forecasting inventories and managing resources are crucial to having a healthy organizations, therefore huge industries such as airlines, use predictive analytics to set ticket prices, and predict the number of guests, or in the case of hospitality, hotels make sure to understand what are the right days and seasons when they get the most customers.
Reducing risk. Companies use predictive models to understand person’s credit scores, which automatically leads to assessing a buyer’s likelihood of default for purchases. Other risk-related uses include insurance claims and collections.
Training and Test data
The customers’ past transactions and current interests, show the training data which tells your model what to look for. At this point, your job is to organize the data into a structure which makes it easily accessible and usable, this could perhaps be a database.
After the first step, one can than start working on data classification, for which part of it is the prediction stage. At this stage, what you are looking for is tests, which show the accuracy of your model. It is important to have historical data of the customers, referred as test data. (hint: do not confuse the term with training data).
When coming up with the values of the two staged explained above, you will be dealing with two possible outcomes: Either you’re satisfied with the accuracy of the model or you aren’t:
· If satisfied, get your model ready to make predictions as part of a production system.
· If not satisfied; retrain your model with a new training dataset.
Predictive Analytics Workflow
1. Access and Explore data: Files, Databases, Sensors
2. Preprocess data: Working with messy data, data reduction, data transformation feature extraction
3. Develop Predictive models: model creation e.g. machine leaning, parameter optimization, model validation
4. Integrate analytics with systems: desktop apps, enterprise scale systems (MATLAB, Excel, Java, C/C++, Python
Challenges of Predictive Analytics
- Relevancy of data
- Discovery of data patterns and relationships for decision making
- Using Hadoop for pre-processing data
- Evaluation of accuracy of the predictive model
- Relate predictive modelling and business concepts for better decision making
- Monitoring the data model while adjusting for necessary changes
One major problem still remains:
Much time and effort needs to be spend in the data preparation (30 to up to 60 percent) when using data from data ware houses. A main reason is that data is often stored without context. The process integrating data from multiple databases become very complex. Modelling the context takes another 20 to 30 percent. The following chart (courtesy OpTier) explains:
In our daily lives every day we ask ourselves how do we boost sales, how do we cut costs, are is crime and human risks fought, how healthcare works? Through Predictive analytics, individuals and organizations are able to predict trends and behavior patterns for unknown future events which highly effects future outcomes. In the meantime using predictive analytics businesses have greater chances to remain competitive, improve customer service and minimize risks. If you are aiming for meaningful and efficient success, understanding and working with predictive analytics in your organization should definitively be a high priority in the list.
What are your thoughts or suggestions?