Applying Neural Networks and Analogous Estimating to Determine the Project Budget

In this episode, Ricardo talks about how neural networks and the analogous estimating can help when determining the project budget, especially when you have access to reliable historical information.

Download (5 MB)

Episode transcript

The transcript is generated automatically by Podscribe, Sonix, Otter and other electronic transcription services.

Hello, everyone. Welcome to the five Minutes PM podcast today. I like to discuss the paper I'm presenting this week at the PMI global Congress, North America here in Orlando in the paper covered the topics Applying Neural Networks and Analogous Estimating to Determine the Project. Budget how this topic came to my mind. I was studying risk management and one of the tools I learned it was exactly about Neural tools for new neural networks. And what does a neural network neural network is a statistical learning model inspired by the way that be a logical nervous system, such as our brain processing formation, the process records one at a time and learning by comparing their classifications with the known ecto classification.

So it's basically, if you try to understand it's I have some historical information based on previous projects, for example, where I know some variables and I know some results, and I use this known information to teach and to help the artificial neural network to understand how is the behavior of that variables. As soon as I teach, I can apply this knowledge to new data, and then this software can extrapolate for me, for example, the project management costs or the project budget.

For example, the neural networks can be applied in, in many other areas, not only in the project management, for example, handwriting recognition, stock market prediction, image, compression sales forecast in do steel process control. So basically what do you do? You create a formula that use the variables, you know, to generate the variable that you don't know, and based on the data you already have this formula that these produce it is more reliable or less reliable.

It's very important. If you have an unreliable data for training, you will produce an unrelated and network, of course. So it's very, very important that you have reliable data. And this is one of the biggest challenges of making this process at work. When we talk about our first step, each two, have the right data set the right reference for training purposes. So you need to have a set of cases of previous projects where independent variables were in place and then output or a dependent variable is calculated.

So let me explain, what are the types of variable? Do you have dependent and independent? What does the dependent and then a variable it's an output. It's a variable that is calculated based on the independent variables. Basically we have two types dependent category. It's not a number. So it's like red, green, blue, high, medium, and low and dependent numeric where the output is a numeric value.

On the independent side, you have also category and numeric. For example, I can create an independent variable in the category area called complexity, Project complexity, and I can say extreme, very high, high, medium, low, very low. And then this independent category will seed and will feed the process. Two calculate dependent one So basically which kind of variable scan we use in the pro's of the environment to teach an also to understand the project complexity.

It can be a category of variable location, Budget X or cost, cost, variance, baseline duration, actual duration, racial, various type of contracting, case of procurement, a number of relevance, stakeholder groups, and then the, what do you do? You start training, you put this on the software in this case, for example, I use it the neuro twos from Palisade. And you do for example, supervisor training, where you know, the inputs and, you know, the output, the actual output, and you used to train to understand if the results are good or bad Minutes if the formula is well-trained it or not, of course, if you have more data, you will create a more reliable training process.

It requires a very complex calculation and a very specific softer you cannot is the basic way to explain this. It's like, do you remember when you study linear regression? It's something similar to that, but in a much more complex environment with a much bigger set of variables. So then you use this to calculate it, to forecast, for example, the project management cost. So you know that some projects in some locations with some complexity, some Budget some duration, some stakeholders, they had some specific product management costs, and you want to estimate then forecast the new project where, you know, the location, you know, the complex, you know, the Budget the duration, but you don't know the project management cost.

And then you use this kind of Neural tools to sort out this. So it's a very useful tool. I love the concept of Swiss army knife for the project manager. I'm not saying that this will solve all of your problems, but its a very helpful to, to determine the project. And also with the development of software tools, this becomes very simple because in the past this was very, very hard because the software tools we're not available, but I want just to leave the final message.

The biggest challenge, all this process is to produce reliable results is to know the quality of the information I have. If I have a weak information or a non reliable information, I can create the whole forecast mechanisms that is completely unreliable and will originate wrong results. So this is the most important topic you need to think about when you work on article official neural networks, but it's a very, Oh nice and reliable to me that I suggest you to take a look on the article on my website and also to take a look on the web.

There's a lot of very nice and great information to learn about artificial. Networks I hope you want, I enjoyed this podcast and see you next week. What if another five minutes before the podcast.

Menu

Resources & Insights

Products & Services

More Resources & Insights

Ricardo Vargas

Podcasts

Applying Neural Networks and Analogous Estimating to Determine the Project Budget

Search in Podcasts

Latest podcast episodes

Book recommendations