The predictions of a binary prediction are negative

Question

I'm working on creating a binary model. I thought everything was working alright, but when I found it weird how often the model was off, but when I tried to adjust the threshold I noticed that nothing changed, so that's when I started to investigate.

I checked my predictedValues of my binary classification and I noticed that most of them were negative.

Here is my model:

        public ITransformer TrainCategorialModel(IEnumerable<TrainingCategorial> trainingData)
    {

        var columnNames = typeof(TrainingCategorial)
            .GetProperties()
            .Where(property => property.DeclaringType != typeof(TrainingCategorial))
            .Select(property => property.Name)
            .ToArray();

        // Check for null values in training data
        
        if (trainingData.Any(item => item == null))
        {
            throw new ArgumentException("Training data contains null values.");
        }

        var pipeline = mLContext.Transforms.Concatenate("Features", columnNames)
            .Append(mLContext.BinaryClassification.Trainers.SdcaNonCalibrated(labelColumnName: "CHPlabels", featureColumnName: "Features"));

        var data = mLContext.Data.LoadFromEnumerable(trainingData);

        var model = pipeline.Fit(data);

        return model;
    }

Where my features are based on the parameter model in another class.

My prediction looks like this:

        public List<bool> PredictCategorialModel(ITransformer model, IEnumerable<PredictionCategorial> input)
    {
        // 4. Transform data
        IDataView testingData = mLContext.Data.LoadFromEnumerable(input);

        // 5. Predict the new values based on the features. 
        List<float> predictedValues = mLContext.Data.CreateEnumerable<BinaryPrediction>(
            model.Transform(testingData), reuseRowObject: false)
            .Select(row => row.LabelPrediction)
            .ToList();

        // Apply a threshold (e.g., 0.5) to convert scores into boolean predictions
        var threshold = 0.3;
            List<bool> predictedLabels = predictedValues.Select(LabelPrediction => LabelPrediction > threshold).ToList();

        return predictedLabels;
    }

I've checked my data and it appears fine. How to fix this?

Update: I think the issue lies in the model, I've tried other ways of creating the predicted boolean, but I get the same errors. I've tried the LightGBM, as I know the model shouldn't be linear, but that created whole new problems (which I also have an unaswered question about). Does anyone know any good ways to check if a model works?