+41 768307656
info@htc-sagl.ch

Tag: supervised learning

Big Data

Vending Machine – Data Analysis – From study to action and how to improve performance

Aug 3, 2023 by Helvetica Technical Consulting 0 Comments #big data, #data scientist, #supervised learning

Data analysis of A vending machine can be very helpful because the information given after trasforming & visusalizing data can enhance logistics, avoid losses and improve performance.

A vending machine is one of those machines installed in shopping mall, offices and stores. They can sell anything which is inside. Any item is stored in a coil and can be bought at a fixed price.

The new models allows to collects usefull data in csv format and then can be manipulated in a way that can give a lot of information like customer profile, spending, preferences and also to discover some correlation between two or more products are sold together.

This study collects data from a single vending machine and try to analyse and search for some correlation between items sold.

Data consist in a single file with 6445 rows and 16 columns. Rows corresponds to a a single operation, from January to August. Most important columns for this study correspond to:

  • Name
  • DateofSale: day,month, day number, year
  • Type of Food: Carbonated, Non-Carbonated, Food, Water
  • Type of Payment: credit card, cash
  • RCoil: coil number of the product
  • RPrice: price of the product in the coil
  • QtySold: quantity sold
  • TransTotal: total amount of the transaction. Normally 1 sold, 1 paid, but can happens that more than item can be sold

Preprocessing

Data is loaded as follow, removing unnecessary fields from raw data:
After cleaning and transforming data, the following table shows the entire dataset consisting of 6445 rows and 10 columns.

The first thing to do, is a preliminary calculation to see which categories are present in the dataset.

We can see that the 2 most important categories are food and carbonated drinks, which correspond to 78% of total transactions in 8 months of sampling. In the following sections we will go deep into data analytics

Carbonated

The following table corresponds Carbonated products and quantity sold, sorted from hightest to lowest:

The first 5 positions, corresponding to 37% of the types of carbonated drink, sold 1431

Food

The following table corresponds Food products and quantity sold, sorted from hightest to lowest

In case of food, the first 5 position covers only 23% of the total quantity sold, in addition to this the number of categories/brands is 7 times bigger than carbonated drinks. This creates a spread in the sales because the user/client has more types to choose. The above short section shown the data extracted from the main dataset that is usefull to provide an indication of trending products. The information given is without any statistical inference, but merely data extracted, loaded and transformed (ELT).

Monthly sales

If you want to see the overall study and discover if there is a correlation between a carbonated drink is sold with food, you can find it below

Machine Learning

Supervised, Unsupervised & Reinforced Learning, a quick intro!

Mar 1, 2023 by Helvetica Technical Consulting 0 Comments #machine learning, #predictive maintenance, #reinforced learning, #supervised learning, #unsupervised learning

In the field of predictive maintenance for rotating equipment, machine learning algorithms can be classified into three categories: supervised learning, unsupervised learning, and reinforced learning. Each of these approaches has its strengths and weaknesses, and choosing the right approach depends on the nature of the problem at hand. In this essay, we will explore the differences between these approaches and their applications in the context of predictive maintenance for rotating equipment.

Supervised Learning

Supervised learning involves training a model on labeled data, where both the input data and the desired output are provided. The goal is to learn a function that can predict the output for new, unseen input data. In the context of predictive maintenance for rotating equipment, supervised learning can be used to predict the remaining useful life of a machine or to detect anomalies that may indicate the onset of a fault.

One common application of supervised learning in predictive maintenance is to analyze vibration data from rotating machinery. By training a model on labeled data that indicates when a fault occurred and the corresponding vibration patterns, the algorithm can learn to identify these patterns in real-time data and predict potential faults before they occur.

Unsupervised Learning

Unsupervised learning involves training a model on unlabeled data, where the input data is provided without any corresponding output. The goal is to find patterns or structures in the data that can be used to make predictions or identify anomalies. In the context of predictive maintenance for rotating equipment, unsupervised learning can be used to identify patterns or clusters in sensor data that may indicate the presence of a fault.

One common application of unsupervised learning in predictive maintenance is to use clustering algorithms to group similar data points together. By analyzing the clusters, it may be possible to identify patterns that are indicative of a specific type of fault or to detect anomalies that may indicate the onset of a fault.

Reinforced Learning

Reinforcement learning involves training a model to make decisions based on feedback from the environment. The goal is to learn a policy that maximizes a reward signal over time. In the context of predictive maintenance for rotating equipment, reinforced learning can be used to develop maintenance schedules that minimize downtime and reduce costs.

One common application of reinforced learning in predictive maintenance is to use a model to determine when maintenance should be performed based on the condition of the machine and the cost of downtime. By learning a policy that balances the cost of maintenance with the cost of downtime, it may be possible to develop a more efficient maintenance schedule that reduces costs and increases efficiency.

Choosing the Right Approach

The choice of machine learning approach depends on the nature of the problem at hand. Supervised learning is best suited for problems where labeled data is available, and the goal is to predict an output for new, unseen data. Unsupervised learning is best suited for problems where the data is not labeled, and the goal is to identify patterns or anomalies in the data. Reinforced learning is best suited for problems where the goal is to develop a policy that maximizes a reward signal over time.

In the context of predictive maintenance for rotating equipment, a combination of these approaches may be used to develop a comprehensive predictive maintenance strategy. For example, supervised learning can be used to predict the remaining useful life of a machine, unsupervised learning can be used to identify patterns or clusters in sensor data, and reinforced learning can be used to develop a maintenance schedule that balances the cost of maintenance with the cost of downtime.

Conclusion

In conclusion, machine learning algorithms can be classified into three categories: supervised learning, unsupervised learning, and reinforced learning. Each of these approaches has its strengths and weaknesses, and choosing the right approach depends on the nature of the problem at hand. In the context of predictive maintenance for rotating equipment, a combination of these approaches may be used to develop a comprehensive predictive maintenance strategy that minimizes downtime, reduces costs