Man&Amp;Machine Consulting}

Man&Machine Consulting

by

manmachine

Leider ist der Eintrag nur auf Amerikanisches Englisch verfgbar. Der Inhalt wird unten in einer verfgbaren Sprache angezeigt. Klicken Sie auf den Link, um die aktuelle Sprache zu ndern.

Machine Learning is the most important field of artificial intelligence nowadays and it will soon be ubiquitously used where a reasonably large amount of data is available. For starters, Machine Learning is the use of algorithms that learn from data and make informed predictions and decisions without being explicitly programmed. It can be simply any data that belongs to the context of the desired use case.

Applications are already in everyday lives. The recommenders on Netflix, YouTube or Amazon that are able to determine correctly (at least often) what you want to view or buy. The home assistants like Amazon Echo or Google Home that can plan your day or play some slow music when youre emotionally down.

Dont expect quality insights out of poor data (aka garbage in, garbage out)

Machine Learning applications depend on the relevance and quality of the data that is analyzed by the algorithms. If the data is inaccurate or irrelevant, the predictions and decisions made will be off the mark:

If you access YouTube using your friends laptop, the recommendation will still give your friends playlist unless you sign in

In order to detect fraud, it is paramount that the historical data used is about correct transactions. If fraud data is the norm, the decision made will be incorrect

Think about a machine learning software that is used to provide the best possible pay for employees. Data showing some employees earning more than the CEO should be filtered out first

Dont expect something that isnt there

For successful operation of a Machine Learning application, the dataset has to be huge enough to make any kind of intelligent prediction. For instance, in home assistant software, the data collected in one week may not provide enough characteristics of the user to provide decisions. The assistant has to learn over several months to understand the repetitive patterns of the user and use that knowledge to predict what the user might need in the coming days.

Another limitation of Machine Learning is that it can only make predictions based on the available data. It can never make up something that the data have not alluded. In a smart program advising farmers on the most convenient crops to plant in a certain area, the program cannot provide any conclusive prediction if the historical data of the climatic conditions and soil type is unavailable.

Dont expect there will be no hard work

Building a quality machine learning application is generally a complex task but there is one aspect that is the most underestimated: Data preparation. Data may be presented in various forms. Some may be structured (like databases or Excel sheets). Other forms may be more or less unstructured (like logfiles, documents or images).

Necessary tasks are e.g. formatting, cleaning and sampling. In formatting, the available data is converted into the required format. The data might be required in a single flat file while the available data may be in several database tables. Or date fields might be in MM/DD/YY format but you need the day of week. Cleaning involves fixing of missing data and removal of redundant and unrelated data. Sampling involves using a small subset of a large amount of data because running the algorithms over the whole dataset is uneconomical and might take a lot of time.

Expect tangible results (if everything is done right)

In a business setup, the use of Machine Learning in various aspects can provide a huge boost:

Management can use data to predict the expected rate of growth in the coming months, e.g. to increase inventories or staff

The finance department can employ Machine Learning techniques to track the number of sales and spot anomalies

The marketing department can use campaign results data to fine-tune their segmentation and generate the next best offer for a client

Digital Transformation

Article Source:

eArticlesOnline.com}