Data analytics and machine learning in Manufacturing

← Go back to overview

Solution Accelerators for Manufacturing

Based on best-practices from our work with the leading brands, we’ve developed solution accelerators for common data analytics and machine learning use cases to save weeks or months of development time for your data engineers and data scientists.

Demand Forecasting with Causals

The growth of e-commerce, volatility with suppliers, and the risk of global pandemics has shocked and accelerated the demands on supply chains. Companies have found existing models and approaches to predicting demand and managing inventory insufficient for the new normal in retail. A company may have run weekly or monthly aggregate forecasts with limited data sets in the past, but competing in the era of e-commerce where consumers can easily switch stores requires that companies have the ability to predict demand for a SKU at a day and store level.


Time-series Forecasting

Improving the speed and accuracy of time series analyses in order to better forecast demand for products and services is critical to retailers’ success. In this notebook we discuss the importance of time series forecasting, visualize some sample time series data, then build a simple model to show the use of Facebook Prophet. Once you’re comfortable building a single model, we’ll combine Prophet with the magic of Apache Spark™ to show you how to train hundreds of models at once, allowing us to create precise forecasts for each individual product-store combination at a level of granularity rarely achieved until now.

Safety Stock

Natural disasters, pandemics, societal unrest and other factors have all recently caused disruptions to our global supply chains. Ensuring that we have enough product to serve demand, while not carrying too much inventory is a key challenge for every business. This solution provides a modern way of helping retailers and manufacturers identify the optimal safety stock to carry to prevent business disruption while freeing working capital.

ML-based Item Matching

How do manufacturers understand what inventory they have on hand around the globe across hundreds of thousands or possibly millions of parts, where local teams could have different item descriptions across internal systems? Or how can a manufacturer resolve differences between their product definitions and those product descriptions across dozens of retail partners? This solution uses machine learning to evolve rules-based and probabilistic (fuzzy) matching techniques for effective product matching on imperfect data.

Scaling ML Models for IoT

Per addestrare modelli di machine learning su dati in tempo reale provenienti da sensori IoT, alcuni casi d'uso richiedono che ogni dispositivo connesso abbia un proprio modello specifico, poiché molti algoritmi elementari di machine learning spesso superano le prestazioni di un singolo modello complesso. Questo può tuttavia generare una quantità di dati IoT e per dispositivo ingestibile per una singola macchina, mentre i dati di ogni dispositivo possono essere gestiti su una macchina. Inoltre, il team di data science sta implementando l'utilizzo di librerie a nodo singolo, come sklearn e pandas, in modo da ridurre gli ostacoli nella distribuzione del loro proof-of-concept con macchina singola. In questo blog spieghiamo come risolvere questo problema con due diversi schemi per ogni dispositivo IoT: Model Training e Model Scoring.

Manutenzione predittiva

La manutenzione di apparecchiature come i compressori è estremamente complessa perché vengono impiegati in svariati tipi di impianti, dalle piccole perforatrici alle piattaforme in alto mare, sono distribuiti in tutto il mondo e generano terabyte di dati ogni giorno. Il guasto di uno solo di questi compressori può costare milioni di dollari in produzione persa ogni giorno. Questa soluzione mostra come costruire una pipeline di dati predittiva completa, in grado di alimentare un database in tempo reale per mantenere le mappature di parti e sensori dell'impianto, supportare un'applicazione continua che elabora una quantità enorme di dati telemetrici, e aiutare a prevedere i guasti del compressore grazie a questi set di dati.

Pronto per cominciare?