There are four types of Analytics:
Prescriptive analytics: What should I do about it?
Prescriptive analytics is about using data and analytics to improve decisions and therefore the effectiveness of actions.Prescriptive analytics is related to both Descriptive and Predictive analytics. While Descriptive analytics aims to provide insight into what has happened and Predictive analytics helps model and forecast what might happen, Prescriptive analytics seeks to determine the best solution or outcome among various choices, given the known parameters.
“Any combination of analytics, math, experiments, simulation, and/or artificial intelligence used to improve the effectiveness of decisions made by humans or by decision logic embedded in applications.”These analytics go beyond descriptive and predictive analytics by recommending one or more possible courses of action. Essentially they predict multiple futures and allow companies to assess a number of possible outcomes based upon their actions. Prescriptive analytics use a combination of techniques and tools such as business rules, algorithms, machine learning and computational modelling procedures.
Prescriptive analytics can also suggest decision options for how to take advantage of a future opportunity or mitigate a future risk, and illustrate the implications of each decision option. In practice, prescriptive analytics can continually and automatically process new data to improve the accuracy of predictions and provide better decision options.
Prescriptive analytics can be used in two ways:
Inform decision logic with analytics: Decision logic needs data as an input to make the decision. The veracity and timeliness of data will insure that the decision logic will operate as expected. It doesn’t matter if the decision logic is that of a person or embedded in an application — in both cases, prescriptive analytics provides the input to the process. Prescriptive analytics can be as simple as aggregate analytics about how much a customer spent on products last month or as sophisticated as a predictive model that predicts the next best offer to a customer. The decision logic may even include an optimization model to determine how much, if any, discount to offer to the customer.
Evolve decision logic: Decision logic must evolve to improve or maintain its effectiveness. In some cases, decision logic itself may be flawed or degrade over time. Measuring and analyzing the effectiveness or ineffectiveness of enterprises decisions allows developers to refine or redo decision logic to make it even better. It can be as simple as marketing managers reviewing email conversion rates and adjusting the decision logic to target an additional audience. Alternatively, it can be as sophisticated as embedding a machine learning model in the decision logic for an email marketing campaign to automatically adjust what content is sent to target audiences.
Different technologies of Prescriptive analytics to create action:
Use of Prescriptive Analytics in ThingWorx Analytics:
Thing Optimizer: Thing Optimizer functionality provides the prescriptive scoring and optimization capabilities of ThingWorx Analytics.
While predictive scoring allows you to make predictions about future outcomes, prescriptive scoring allows you to see how certain changes might affect future outcomes. After you have generated a prediction model (also called training a model), you can modify the prescriptive attributes in your data (those attributes marked as levers) to alter the predictions. The prescriptive scoring process evaluates each lever attribute, and returns an optimal value for that feature, depending on whether you want to minimize or maximize the goal variable.
Prescriptive scoring results include both an original score (the score before any lever attributes are changed) and an optimized score (the score after optimal values are applied to the lever attributes). In addition, for each attribute identified in your data as a lever, original and optimal values are included in the prescriptive scoring results.
How to Access Thing Optimizer Functionality:
ThingWorx Analytics prescriptive scoring can only be accessed via the REST API Service. Using a REST client, you can access the Scoring service which includes a series of API endpoints to submit scoring requests, retrieve results, list jobs, and more. Requires installation of the ThingWorx Analytics Server.
How to avoid mistakes - Below are some common mistakes while doing Prescriptive analytics:
Image source: Wikipedia, Content: go.forrester.com(Partially)
I am confused about the scoring part. what does the value of prescriptive score mean? I saw decimals eg 0.4567 for both optimized score and original score in (prescriptive realtime score). What is the meaning of eg 0.4567?
In predictive scoring(realtime score), i can see the 'goal' which is the prediction and 'goal_mo' which is also 0.4567. What is 'goal_mo' and is this the same value shown in prescriptive thing?
I have a few question on this
Is this _mo in predictive thing used in the prescriptive thing as well?
In my case, the goal i want to maximise is this field called 'Goal' which is strings : 'good', 'normal', 'bad'
Is the process like this?
correct me if i am wrong:
1: The inputs eg vibration/ temperature readings are scaled to 0,1 to train
2: During testing/scoring phase, the inputs are scaled too.
3. The output aka _mo is being output as between 0-1 and scaled back to original scale.
4. using the scaled output the model will map to the corresponding strings
But in this case how will the model know what the definition of good, normal, bad as i nvr specifiy what is good?
In this case how does optimization even happen?
But that does not seem to be the case when i tried using the real time scoring in prescriptive thing.
I get the result as shown below with no categories.
I see. The link seems helpful so i will most probably try it.
I am planing to work with time series datasets so the question is:
Does thingworx prescriptive thing work with time series datasets?
I was reading up on how prescriptive analytics can in real time suggesting actions to benefit from the predictions and show the implications of each decision option on the end goal eg. profit so i wonder if thingworx has such functionality.
Interestingly, i realize that you can get the the following error for wrong range input. How does the TWA determine the correct boundary for this dataset?
Failed to score: Invalid range specified for the continuous field [Cement].
Acceptable bound(s): [102.0,540.0]
is it possible to briefly explain how the prescriptive analytics finds the values to maximise or minimise the goal? I got a little confused as the model i trained on is a mashup of neural network and some other algorithm. From what i understand, it acts like a black box with a lot of functions in the background so it seem to be impossible to find the global max or min.
Does Thingworx(prescriptive analytics) use brute force on all the possible values of the chosen levers to find the most optimal value?
During pre-sales I often get asked by customers about the details on how TWA works under-the-hood. Is it documented somewhere, besides the info in help center?
I mean, I can explain DS concepts to people, but the people asking such questions already have some DS expertise and they want to know what approaches are implemented in TW Analytics to try to estimate quality of the models and predictions they'll be getting, otherwise they're reluctant to trust the "black box" software.
Even if we persuate the decision maker and win the deal, having a sceptical DS guy who will be implementing the solution will not be benefical to the project.
Yes, predictions themselves is just one piece of TW and TWA functionality and the product has other major benefits which I communicate during pre-sales, but I couldn't find a good way to handle such kind of questions.
Personally, as a user i think that Thingworx is a great platform for the non technical to use it as 'black box' tool. For DS people, it indeed serves as a white box. However, there will bound be users that may be unfamiliar with some of the newer concepts like prescriptive analytics which i personally feel that there should be clearer explanations of the features for non DS and DS people. (E.g. a tutorial which they can follow to understand. I feel that the concrete example was great!)
For DS people, a separate section that goes more in depth into the features like the meaning behind the numbers/signels or even a brief explanation on what is going on behind the scenes will also be beneficial for both the sales team and end user as well.
Thank you for the prompt response, @Rocko
Special kudos for mentioning the type of math behind profiles - this is the kind of information I'm looking for. I actually thought it was based on frequent pattern mining, something like fp growth or apriori.
And thank you for mentioning the training - I've checked the materials from it and there is indeed a lot of information to which I either didn't pay attention back then or which wasn't emphasized during the training. Even beam search is there.