NILG.AI

Unlocking business capabilities using data intelligence

[NILG.AI @nilg_ai]

Anticipating actions is one way to use future predictions to transform your business. Predict what would happen under different realities and choose the one that puts you in the best situation #AI #ML #data https://t.co/zQTJe2d3Se

[NILG.AI @nilg_ai]

Property managers can use AI to avoid involuntary tenant turnover by proposing the right terms and payment plans, forecasting repair demands, and optimizing repairs and scheduling. #AI #ML #data https://t.co/abBYWGzOFF

[NILG.AI @nilg_ai]

Remember, during ideation states of AI projects, exploring a diverse set of options is more important than choosing. You want to create options. #AI #ML #data https://t.co/qj5TFpuzPY

[NILG.AI @nilg_ai]

Models with high predictive performance but low decision support can be improved by focusing on elasticity and ensuring they are responsive to changes in control variables. #AI #ML #data https://t.co/8jY0yWafLJ

[NILG.AI @nilg_ai]

Short-term predictions tend to be more accurate and require less historical data, while long-term predictions increase actionability but decrease predictive power and require more data. #AI #ML #data https://t.co/dZo34LLg0u

[NILG.AI @nilg_ai]

In credit scoring, approving loans across the spectrum allows us to improve over time. We need to do this in a controlled manner to maintain a profitable business while still keeping a learning population to mitigate and remove sources of bias.

[NILG.AI @nilg_ai]

AI can optimize the Real Estate construction process by automating design, digitizing blueprints, and ensuring compliance with safety procedures. #AI #ML #data https://t.co/abBYWGzOFF

[NILG.AI @nilg_ai]

Business-centric AI means fitting AI into your business pipeline, processes, and culture. It's about how AI feeds your value proposition and how it might change it. #AI #ML #data https://t.co/GbTMERpc7I

[NILG.AI @nilg_ai]

The main problem with AI in hospital management is conflicting objectives: making hospitals more efficient while also making them more human. #AI #ML #data https://t.co/ODu7hL0in0

[NILG.AI @nilg_ai]

Unifying multiple machine learning models into a single one can lead to better results than having individual models for each segment. #AI #ML #data https://t.co/MnSD7HjkRe

[NILG.AI @nilg_ai]

If your model isn't reacting to control variables, try regularizing it, removing highly correlated variables, or changing the target. #AI #ML #data https://t.co/8jY0yWafLJ

[NILG.AI @nilg_ai]

Storing data in a way that can be used for machine learning models can be challenging. Here are three strategies: 1) Keep the final version 2) Keep a snapshot 3) Store changes. #AI #ML #data https://t.co/M2YqJ3hUdB

[NILG.AI @nilg_ai]

The most critical type of risks are high probability, high loss. These are the risks you need to pay attention to and mitigate. Leverage AI for prevention and recovery. #AI #ML #data https://t.co/KTEcWjEf0Z

[NILG.AI @nilg_ai]

One strategy to get intuition on how much data you need is to look for references from similar problems that have already been solved. #AI #ML #data https://t.co/YNUqaU1gPL

[NILG.AI @nilg_ai]

Reducing manual workloads and error rates are two ways to leverage predictions from the present. By automating processes and lowering failure rates, you can improve your business efficiency and profitability. #AI #ML #data https://t.co/zQTJe2d3Se

[NILG.AI @nilg_ai]

Hire a #datascientist first, then a #dataengineer. This approach will allow you to validate, get some short-term profit, and use this profit to reinvest it in building the right infrastructure. #AI #ML #data https://t.co/TmsHH0Ij90

[NILG.AI @nilg_ai]

Explicit and implicit constraints can limit your model to a certain type of population or demographics. Collecting data from all populations is necessary for a massive expansion. #AI #ML #data https://t.co/i71pj0jIQi

[NILG.AI @nilg_ai]

Building a #dummy model or replicating the current process will give you a new #benchmark to measure performance and gain additional business understanding. Then, go iteratively #data source by data source, feature by feature, and always compare with the baseline.

[NILG.AI @nilg_ai]

If you identify that delayed data is an issue, you can simulate the delay in your feature construction. For example, if sales data is integrated monthly, consider that you can only observe up to last month. #AI #ML #data https://t.co/wjEIzhny5O

[NILG.AI @nilg_ai]

Generative AI models can embed biases found on the internet. Ensure the outcomes align with your company culture and values. #AI #ML #data https://t.co/x6ksA5SM8c